{
  "source": "https://cirunners.org",
  "license": "CC BY 4.0",
  "reviewed": "2026-11-22",
  "disclaimer": "General information, provided as is and without warranty. Not professional advice. Product details change; confirm them with each provider.",
  "terms": "https://cirunners.org/terms/",
  "providers": [
    {
      "id": "macstadium-orka",
      "name": "Orka",
      "company": "MacStadium, Inc.",
      "hq": {
        "city": "Atlanta",
        "region": "Georgia",
        "country": "US",
        "source": "https://www.macstadium.com/pricing"
      },
      "kind": "byoc",
      "category": "macos",
      "website": "https://www.macstadium.com/orka",
      "docs": "https://docs.macstadium.com",
      "pricing_url": "https://www.macstadium.com/pricing",
      "github": "https://github.com/macstadium/orka-github-actions-integration",
      "tagline": "Kubernetes-based orchestration of macOS VMs on Apple silicon, run in MacStadium data centers, on AWS EC2 Mac or on your own Macs.",
      "ci_systems": [
        "GitHub Actions",
        "GitLab CI",
        "Buildkite",
        "Jenkins",
        "Azure Pipelines",
        "TeamCity"
      ],
      "os": [
        "macOS"
      ],
      "arch": [
        "arm64"
      ],
      "isolation": "Jobs run in macOS virtual machines on Apple silicon Mac nodes; the GitHub runner creates a VM per job and tears it down afterwards, while the GitLab, Buildkite and Jenkins integrations offer both permanent and per-job (ephemeral) agents",
      "isolation_short": "macOS VM",
      "isolation_class": "vm",
      "ephemeral": null,
      "open_source": false,
      "license": null,
      "self_host": true,
      "byoc": true,
      "clouds": [
        "MacStadium Cloud",
        "AWS",
        "Kubernetes",
        "Bare metal"
      ],
      "gpu": null,
      "gpu_detail": null,
      "hardware": "MacStadium lists hosted Mac mini (M2, M2 Pro, M4, M4 Pro) and Mac Studio (M1 Max, M2 Ultra) configurations; Orka 4.0 runs on Apple silicon nodes only.",
      "performance": null,
      "sizes": "CPU and memory are set per VM config; listed hosted Macs range from an M2 Mac mini (8 cores, 8 GB) to an M2 Ultra Mac Studio (24 cores, 128 GB)",
      "max_size_short": null,
      "cache": "Orka images are OCI images pulled from a registry, and the docs describe pre-caching images on nodes. For GitLab, cache restoration needs the gitlab-runner binary inside the VM image.",
      "docker_layer_cache": null,
      "concurrency": "Bounded by the Mac nodes in the cluster; the GitHub runner reports up to MAX_RUNNERS (default 9000) and GitLab concurrency is set in config.toml",
      "job_limits": null,
      "static_ips": null,
      "ssh_access": true,
      "observability": "The Orka GitHub runner can expose Prometheus metrics, and Orka is Kubernetes-native so clusters can be monitored with Kubernetes tools.",
      "setup": "Obtain an Orka cluster (MacStadium-hosted, on AWS or on-prem), then run the Orka GitHub runner container with a GitHub App and an Orka service account token. GitLab, Buildkite, Jenkins and TeamCity use MacStadium’s scripts or plugins.",
      "runs_on": "my-github-runner",
      "pricing": "Orka for DevOps pricing is not published (contact sales; Orka is also sold through AWS Marketplace). Individual hosted Macs are listed monthly, from $109 (M2 Mac mini, 8 GB) and $149 (M4 Mac mini, 16 GB) to $449 (M2 Ultra Mac Studio, 128 GB).",
      "price_short": "From $109/mo per hosted Mac mini (Orka by quote)",
      "free_tier": "No free trial on individual hosted Macs; MacStadium says a proof of concept may be possible for large migrations",
      "compliance": "ISO 27001, 27017 and 27018; SOC 1 and SOC 2 Type 2 (as stated by MacStadium)",
      "regions": "MacStadium data centers in Atlanta, Las Vegas and Dublin; or your own AWS account or premises",
      "best_for": "Teams that want a dedicated, orchestrated fleet of Macs for ephemeral macOS CI across several CI systems",
      "overview": [
        "**Orka** is MacStadium’s virtualization and orchestration platform for macOS workloads. A Kubernetes control plane schedules macOS VMs, built from OCI images, onto Apple silicon Mac nodes, and teams drive it through the `orka` CLI, a REST API or a web UI. An Orka cluster can run in MacStadium’s data centers, on [AWS EC2 Mac](https://docs.macstadium.com/orka/orka-on-aws-and-on-prem/orka-on-aws-getting-started), or on Macs you own.",
        "For CI, MacStadium publishes integrations rather than a hosted runner label. The [Orka GitHub runner](https://github.com/macstadium/orka-github-actions-integration) is a container that uses GitHub runner scale sets to create a VM per job and delete it afterwards; GitLab, Buildkite and Jenkins integrations support permanent or per-job agents. Orka 4.0, announced on October 28, 2026, runs on Apple silicon only and can pair Android emulators with each macOS VM."
      ],
      "notable": [
        "October 28, 2026: MacStadium announced Orka 4.0, which adds Android emulators alongside macOS VMs and runs on Apple silicon only",
        "Orka 3.6 is the last release to support Intel; MacStadium says it will support it through March 2028",
        "May 6, 2024: the older GitHub Actions guide in orka-integrations was deprecated in favour of the Orka GitHub runner"
      ],
      "pros": [
        "You want per-job macOS VMs on Macs dedicated to your team rather than a shared runner pool",
        "You use more than one CI system: MacStadium documents GitHub Actions, GitLab, Buildkite, Jenkins, TeamCity and Azure Pipelines",
        "You need the same platform in a vendor data center, in your AWS account or on-premises",
        "You want Android emulators next to macOS VMs on the same Mac fleet (Orka 4.0)",
        "You need a vendor that states ISO 27001 and SOC 2 Type 2 coverage"
      ],
      "cons": [
        "Orka for DevOps pricing is not published; MacStadium directs buyers to sales",
        "MacStadium states it does not offer free trials on individual hosted Macs",
        "Orka 4.0 does not support Intel Macs; Intel workloads must stay on Orka 3.6",
        "You run the runner container or CI plugin yourself, with network access to the cluster (the docs note a VPN may be required)"
      ],
      "verdict": "Orka is commercial macOS VM orchestration for teams that want a dedicated Mac fleet, hosted by MacStadium or self-run, wired into the CI system they already use.",
      "use_cases": {
        "macos-ios": "Orka provisions macOS VMs on Apple silicon for Xcode builds and tests, with documented integrations for common CI systems.",
        "byoc": "Orka can be deployed on AWS EC2 Mac instances in your own account or on Mac hardware you own.",
        "gitlab": "MacStadium documents GitLab shell executors and a custom executor that creates an Orka VM per pipeline."
      },
      "peers": [
        "tart",
        "getmac",
        "bitrise-runners",
        "github-hosted-runners",
        "cirrus-runners",
        "namespace",
        "warpbuild",
        "buildkite-hosted-agents"
      ],
      "sources": [
        "https://www.macstadium.com/orka",
        "https://www.macstadium.com/pricing",
        "https://www.macstadium.com/company",
        "https://www.macstadium.com/datacenters",
        "https://www.macstadium.com/blog/orka-4-0-build-and-test-ios-and-android-on-one-apple-silicon",
        "https://docs.macstadium.com/orka/orka-overview/orka-overview",
        "https://docs.macstadium.com/orka/orka-overview/how-orka-manages-vms",
        "https://docs.macstadium.com/orka/orka-devops-integrations/github-actions",
        "https://docs.macstadium.com/orka/orka-devops-integrations/gitlab",
        "https://docs.macstadium.com/orka/orka-devops-integrations/buildkite",
        "https://docs.macstadium.com/orka/orka-devops-integrations/jenkins",
        "https://docs.macstadium.com/orka/orka-upgrades-and-release-notes/orka-release-notes",
        "https://docs.macstadium.com/llms.txt",
        "https://github.com/macstadium/orka-github-actions-integration",
        "https://github.com/macstadium/orka-integrations/tree/master/GitHub"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "tart",
      "name": "Tart",
      "company": "OpenAI",
      "hq": {
        "city": "San Francisco",
        "region": "California",
        "country": "US",
        "source": "https://openai.com/policies/privacy-policy/"
      },
      "kind": "byoc",
      "category": "macos",
      "website": "https://tart.run",
      "docs": "https://tart.run/quick-start/",
      "pricing_url": null,
      "github": "https://github.com/openai/tart",
      "tagline": "Virtualization toolset for macOS and Linux VMs on Apple silicon, with OCI-registry images and the Orchard orchestrator for clusters of Macs.",
      "ci_systems": [
        "GitLab CI",
        "Buildkite",
        "GitHub Actions"
      ],
      "os": [
        "macOS",
        "Linux"
      ],
      "arch": [
        "arm64"
      ],
      "isolation": "Jobs run in macOS or Linux virtual machines created with Apple’s Virtualization.Framework; the documented GitLab executor and Buildkite plugin run each job in an isolated ephemeral VM cloned from an image",
      "isolation_short": "Apple Virtualization VM",
      "isolation_class": "vm",
      "ephemeral": true,
      "open_source": false,
      "license": "FSL-1.1-ALv2",
      "self_host": true,
      "byoc": true,
      "clouds": [
        "Bare metal"
      ],
      "gpu": null,
      "gpu_detail": null,
      "hardware": "Runs on Apple silicon Macs with macOS 13.0 (Ventura) or later; Orchard workers run on macOS and the Orchard controller can also run on Linux or Kubernetes.",
      "performance": "The Tart README says Apple’s Virtualization.Framework gives near-native performance (project claim)",
      "sizes": null,
      "max_size_short": null,
      "cache": "VM images are pushed to and pulled from OCI-compatible container registries. The GitLab executor example sets builds and cache directories inside the VM.",
      "docker_layer_cache": null,
      "concurrency": "The Orchard announcement says a single host allows a maximum of two virtual machines at a time; Orchard schedules VMs across a cluster of hosts",
      "job_limits": null,
      "static_ips": null,
      "ssh_access": true,
      "observability": "Orchard controller and workers emit OpenTelemetry metrics.",
      "setup": "Install with Homebrew (brew install openai/tools/tart) and clone a macOS image from a registry. Add the GitLab Tart Executor to a self-hosted GitLab Runner or the Tart plugin to a Buildkite pipeline; Orchard adds a controller and workers for several hosts.",
      "runs_on": "image: ghcr.io/cirruslabs/macos-tahoe-base:latest",
      "pricing": "No licence fees: Cirrus Labs said on April 7, 2026 that it had stopped charging licensing fees for Tart, Vetu and Orchard. You supply and pay for the Mac hardware.",
      "price_short": null,
      "free_tier": "Free to use under FSL-1.1-ALv2; the earlier paid tiers no longer apply, per tart.run",
      "compliance": null,
      "regions": "Wherever your Apple silicon hosts run",
      "best_for": "Teams that run their own Apple silicon Macs and want scriptable, image-based macOS VMs for GitLab CI or Buildkite jobs",
      "overview": [
        "**Tart** is a command-line toolset for building, running and managing macOS and Linux virtual machines on Apple silicon, using Apple’s Virtualization.Framework. VM images are stored in OCI-compatible container registries, so a host clones an image much as it would pull a container. [Orchard](https://github.com/openai/orchard), its companion orchestrator, runs Tart VMs across a cluster of Macs through a controller, workers and a REST API.",
        "Tart was created by Cirrus Labs, which announced on April 7, 2026 that it had agreed to join OpenAI. The repositories now sit under the `openai` GitHub organisation and, per tart.run, Tart has been maintained by OpenAI under **FSL-1.1-ALv2** since June 5, 2026, with no licence fees. Documented CI integrations are a GitLab Runner custom executor and a Buildkite plugin; the GitHub Actions entry on tart.run points to the Cirrus Runners service."
      ],
      "notable": [
        "April 7, 2026: Cirrus Labs announced an agreement to join OpenAI as part of its Agent Infrastructure team and said it had stopped charging licensing fees for Tart, Vetu and Orchard",
        "June 5, 2026: Tart and Orchard were relicensed under FSL-1.1-ALv2; tart.run says Tart is maintained by OpenAI as of that date",
        "February 11, 2023: Tart moved from AGPL-3.0 to Fair Source 100, a change tart.run now marks as historical",
        "November 17, 2026: Tart 2.42.0 released on GitHub"
      ],
      "pros": [
        "You already own or rent Apple silicon Macs and want to run CI jobs in disposable VMs on them",
        "You want VM images versioned and distributed through an OCI container registry",
        "You use GitLab CI or Buildkite, for which Tart documents an executor and a plugin",
        "You need to schedule VMs across many Macs: Orchard provides a controller, workers and a REST API",
        "You want to read the source: both Tart and Orchard are public on GitHub"
      ],
      "cons": [
        "FSL-1.1-ALv2 is a source-available licence that excludes Competing Use; each version becomes Apache-2.0 on the second anniversary of its release",
        "A single host allows a maximum of two virtual machines at a time, per the Orchard announcement",
        "Only one Orchard controller instance is supported at the moment, per the Orchard docs",
        "The GitHub Actions integration listed on tart.run is Cirrus Runners, which Cirrus Labs said on April 7, 2026 is no longer accepting new customers"
      ],
      "verdict": "Tart is a source-available VM toolset, now maintained by OpenAI, for teams that operate their own Apple silicon Macs and want image-based ephemeral macOS VMs for CI.",
      "use_cases": {
        "macos-ios": "Tart runs macOS VMs on Apple silicon from prebuilt images that include Xcode, for use in CI and other automation.",
        "byoc": "Tart and Orchard run entirely on Mac hardware you operate.",
        "gitlab": "The GitLab Tart Executor runs GitLab jobs in ephemeral Tart VMs through GitLab Runner’s custom executor.",
        "arm64": "Tart also runs arm64 Linux VMs (Ubuntu, Debian, Fedora images) on Apple silicon hosts.",
        "open-source": "The source is public and self-hosted, under the source-available FSL-1.1-ALv2 licence rather than an OSI-approved one."
      },
      "peers": [
        "macstadium-orka",
        "cirrus-runners",
        "getmac",
        "gitlab-runner",
        "actions-runner-controller",
        "buildkite-elastic-ci-stack",
        "github-hosted-runners"
      ],
      "sources": [
        "https://cirruslabs.org/",
        "https://tart.run/",
        "https://tart.run/quick-start/",
        "https://tart.run/blog/2023/02/11/changing-tart-license/",
        "https://tart.run/blog/2023/04/25/announcing-orchard-orchestration-for-managing-macos-virtual-machines-at-scale/",
        "https://tart.run/integrations/gitlab-runner/",
        "https://tart.run/integrations/buildkite/",
        "https://tart.run/orchard/architecture-and-security/",
        "https://tart.run/orchard/deploying-controller/",
        "https://tart.run/orchard/managing-cluster/",
        "https://github.com/openai/tart",
        "https://github.com/openai/tart/blob/main/LICENSE",
        "https://github.com/openai/tart/releases",
        "https://github.com/openai/orchard",
        "https://openai.com/policies/privacy-policy/"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "getmac",
      "name": "GetMac",
      "company": "Innospike, UAB",
      "hq": {
        "city": "Kaunas",
        "region": null,
        "country": "LT",
        "source": "https://getmac.io/contact"
      },
      "kind": "managed",
      "category": "macos",
      "website": "https://getmac.io",
      "docs": "https://getmac.io/docs",
      "pricing_url": "https://getmac.io/pricing",
      "github": "https://github.com/getmac-io",
      "tagline": "Managed macOS runners on vendor-owned Apple silicon M4 hardware in Lithuania, sold as monthly plans of included compute minutes.",
      "ci_systems": [
        "GitHub Actions",
        "GitLab CI",
        "Buildkite"
      ],
      "os": [
        "macOS"
      ],
      "arch": [
        "arm64"
      ],
      "isolation": "Each runner is a hardware-virtualised macOS guest created through Apple’s Virtualization framework that serves exactly one job and is destroyed when the job ends",
      "isolation_short": "macOS VM",
      "isolation_class": "vm",
      "ephemeral": true,
      "open_source": false,
      "license": null,
      "self_host": false,
      "byoc": false,
      "clouds": null,
      "gpu": null,
      "gpu_detail": null,
      "hardware": "Apple silicon M4 Mac hardware that GetMac says it owns and operates in a Tier III data centre in Lithuania.",
      "performance": "GetMac says its M4 runners are typically 2.5x to 3x faster than GitHub-hosted Intel macos-12 runners (vendor claim)",
      "sizes": "Mac M4 Base: 4 cores, 8 GB (1x minute multiplier). Mac M4 Pro (8 cores, 16 GB, 2x) and Mac M4 Max (12 cores, 32 GB, 4x) are listed as coming soon",
      "max_size_short": null,
      "cache": "GitHub Actions cache and artifact actions work as on GitHub-hosted runners. Because runners are ephemeral, GetMac says caches should be stored with GitHub’s cache actions or external storage.",
      "docker_layer_cache": null,
      "concurrency": "1 concurrent VM on Free and Developer, 3 on Team, 6 on Growth, 10 on Business and on self-serve Enterprise sizes",
      "job_limits": "A CI job can run for up to 6 hours; manually created instances are deleted after 60 minutes",
      "static_ips": null,
      "ssh_access": true,
      "observability": "Dashboard with an activity log and a usage API; incidents are published on a public status page.",
      "setup": "Install the GetMac GitHub App and change runs-on to getmac. GitLab CI uses the getmac-gitlab-executor custom executor and Buildkite a plugin, both with a GetMac API key.",
      "runs_on": "getmac",
      "pricing": "Monthly plans with included compute minutes: Free $0 (100 minutes), Developer $11.99 (1,000), Team $33.99 (3,000), Growth $66.99 (6,000), Business $110.99 (10,000). Self-serve Enterprise sizes cost $0.011 a minute, from $220 a month for 20,000 minutes. Top-up packs start at $10 for 600 minutes ($0.0167 per minute).",
      "price_short": "$0.012/min (macOS M4, Developer plan)",
      "free_tier": "Free plan with 100 compute minutes a month and 1 concurrent VM, no credit card required; the changelog also announces free runners for open source projects",
      "compliance": null,
      "regions": "One documented region, eu-central-ltu-1, in a Tier III data centre in Lithuania",
      "best_for": "Small and mid-sized teams that want low-cost Apple silicon runners for GitHub Actions, GitLab CI or Buildkite and can run in an EU data centre",
      "overview": [
        "**GetMac** is a managed macOS runner service operated by Innospike, UAB, a Lithuanian company. After a team installs its GitHub App and sets `runs-on: getmac`, each job gets a macOS VM created through Apple’s Virtualization framework on M4 hardware; the VM serves one job and is then destroyed. Labels such as `getmac-tahoe` and `getmac-sequoia` pin a macOS and Xcode image.",
        "GetMac says it owns its hardware and runs it in a Tier III data centre in Lithuania rather than on a public cloud. Billing is by plan: a monthly fee buys a pool of compute minutes and a concurrency limit, with prepaid top-up packs for extra minutes. Besides GitHub Actions, it documents a GitLab Runner custom executor and a Buildkite plugin, added in February 2026, plus SSH and remote-display access to instances."
      ],
      "notable": [
        "May 18, 2024: changelog records initial support for GitHub Actions runners",
        "August 15, 2025: free macOS GitHub runners for open source projects introduced",
        "February 20, 2026: GitLab CI and Buildkite runner support added",
        "November 16, 2026: self-serve Enterprise plans, top-up minutes and burst minutes added"
      ],
      "pros": [
        "You want a one-line switch from GitHub-hosted macOS runners: install the GitHub App and set runs-on to getmac",
        "You want each job on a single-use VM that is destroyed afterwards",
        "You need macOS runners for GitLab CI or Buildkite as well as GitHub Actions",
        "You want a free plan to test with: 100 minutes a month without a credit card",
        "You need your CI to run on vendor-owned hardware inside the European Union"
      ],
      "cons": [
        "Self-service plans come without a specific availability commitment, per the terms of service",
        "GetMac’s terms state it does not currently hold ISO 27001 certification or a SOC 2 report",
        "Only the Mac M4 Base type (4 cores, 8 GB) is available; M4 Pro and M4 Max are listed as coming soon",
        "When plan, burst and top-up minutes are used up, new jobs wait in the queue until the next billing period"
      ],
      "verdict": "GetMac is a plan-based managed macOS runner service on M4 hardware in Lithuania, suited to teams that want inexpensive ephemeral runners for GitHub Actions, GitLab CI or Buildkite.",
      "use_cases": {
        "macos-ios": "GetMac runs single-use macOS VMs on M4 hardware with Xcode images selected by runs-on label.",
        "lower-cost": "Plan minutes work out at $0.0111 to $0.0120 a minute, and GetMac says it is up to 70% cheaper than GitHub’s per-minute pricing (vendor claim).",
        "gitlab": "A custom GitLab Runner executor creates a fresh GetMac VM for each GitLab job and deletes it afterwards."
      },
      "peers": [
        "github-hosted-runners",
        "bitrise-runners",
        "macstadium-orka",
        "cirrus-runners",
        "namespace",
        "warpbuild",
        "depot",
        "blacksmith"
      ],
      "sources": [
        "https://getmac.io/",
        "https://getmac.io/pricing",
        "https://getmac.io/faq",
        "https://getmac.io/terms-of-service",
        "https://getmac.io/contact",
        "https://getmac.io/about",
        "https://getmac.io/infrastructure",
        "https://getmac.io/changelog",
        "https://getmac.io/docs/github-runners/getting-started",
        "https://getmac.io/docs/gitlab-runners/getting-started",
        "https://getmac.io/docs/buildkite-runners/getting-started",
        "https://getmac.io/docs/runner-images/workflow-labels",
        "https://getmac.io/docs/billing/top-up-minutes",
        "https://getmac.io/docs/billing/enterprise-plans"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "bitrise-runners",
      "name": "Bitrise Build Hub",
      "company": "Bitrise",
      "hq": null,
      "kind": "managed",
      "category": "macos",
      "website": "https://bitrise.io/platform/build-hub",
      "docs": "https://docs.bitrise.io/en/bitrise-build-hub",
      "pricing_url": "https://bitrise.io/pricing",
      "github": null,
      "tagline": "Bitrise-managed macOS and Linux runners for GitHub Actions on Bitrise-owned Apple silicon and AMD EPYC machines, sold as monthly compute budgets.",
      "ci_systems": [
        "GitHub Actions"
      ],
      "os": [
        "macOS",
        "Linux"
      ],
      "arch": [
        "arm64",
        "x64"
      ],
      "isolation": "Jobs run on virtual machines from pre-warmed pools; with GitHub App authentication a machine is provisioned when a workflow starts and shut down when the build ends, while with a personal access token the requested number of machines is kept running",
      "isolation_short": "Virtual machine",
      "isolation_class": "vm",
      "ephemeral": true,
      "open_source": false,
      "license": null,
      "self_host": false,
      "byoc": null,
      "clouds": null,
      "gpu": null,
      "gpu_detail": null,
      "hardware": "Bitrise-owned Mac minis (M4 and M4 Pro, with M5 Pro and M6 listed on the Enterprise plan) and AMD EPYC Zen 4/Zen 5 Linux machines in data centres that Bitrise operates.",
      "performance": "Bitrise publishes its own benchmark of 2m 27s on a Build Hub M5 Pro machine against 8m 43s on a GitHub-hosted M2 runner for the DuckDuckGo iOS app (vendor claim)",
      "sizes": "Self-serve plans: macOS M4 Large (5 vCPU, 14 GB) and M4 X Large (10 vCPU, 28 GB); Linux from 2Small (2 vCPU, 8 GB) to X Large (16 vCPU, 64 GB). Enterprise lists M4 Pro, M5 Pro and M6 and Linux up to 192 vCPU",
      "max_size_short": "16 vCPU / 64 GB",
      "cache": "A machine pool can route GitHub Actions cache requests to Bitrise infrastructure instead of GitHub’s cache backend. The separately priced Bitrise Build Cache (Gradle, Bazel, Xcode, React Native) needs no token on Build Hub machines.",
      "docker_layer_cache": null,
      "concurrency": "Counted in vCPU: Basic up to 35 macOS and 136 Linux vCPU, Pro up to 90 macOS and 640 Linux vCPU, depending on tier; the trial allows 30 macOS and 48 Linux vCPU",
      "job_limits": null,
      "static_ips": true,
      "ssh_access": true,
      "observability": "The machine pool view shows per-machine status, live stdout and stderr logs and, where enabled, CPU, memory and disk charts.",
      "setup": "Create a machine pool in a Bitrise workspace, install the Bitrise - Build Hub GitHub App (or use a personal access token on Enterprise), then target the pool’s labels in runs-on.",
      "runs_on": "bitrise-runner-my-pool",
      "pricing": "Plans are monthly compute budgets: Basic from $15/month ($13.50 billed annually), Pro from $210/month ($189 billed annually), Enterprise by quote. Machine rates per minute: macOS M4 Large $0.03, M4 XLarge $0.06; Linux 2Small (2 vCPU) $0.00625, Medium (4 vCPU) $0.0125, Large (8 vCPU) $0.025, XLarge (16 vCPU) $0.05. Pay-As-You-Go rates are listed at 2x these.",
      "price_short": "$0.00625/min",
      "free_tier": "14-day free trial with all Basic and Pro machine types, limited to 30 macOS vCPU and 48 Linux vCPU",
      "compliance": "SOC 2 Type II and PCI DSS (as stated by Bitrise)",
      "regions": "US and EU data centres operated by Bitrise",
      "best_for": "Mobile teams on GitHub Actions that want managed Apple silicon and Linux runners with iOS and Android toolchains preinstalled",
      "overview": [
        "**Bitrise Build Hub** runs GitHub Actions jobs on machines that Bitrise, the mobile CI vendor, owns and operates. A team creates a machine pool in a Bitrise workspace, connects it through the Bitrise - Build Hub GitHub App, and points `runs-on` at the pool’s labels; workflow files otherwise stay as they are. Pools use Apple silicon Mac minis or AMD EPYC Linux machines with Bitrise’s mobile build stacks.",
        "Build Hub is sold separately from Bitrise CI, as monthly compute budgets shared between macOS and Linux and drawn down at per-minute machine rates. It works with GitHub Cloud and GitHub Enterprise Server. Pools can send GitHub Actions cache requests to Bitrise’s co-located cache, run a warmup script, and expose SSH or VNC access to a running machine. Bitrise says new Xcode versions are available within 24 hours of release."
      ],
      "notable": [
        "August 25, 2026: Bitrise announced that M5 Pro and M6 Mac mini runners would be added to Build Hub",
        "June 1, 2026: a Bitrise blog post described three Bitrise-operated data centres (two in the US, one in the EU)",
        "September 22, 2026: docs changelog notes that Build Hub runners authenticate to the Bitrise Build Cache without a token"
      ],
      "pros": [
        "You build iOS and Android apps on GitHub Actions and want macOS and Linux runners from one vendor",
        "You want stacks with Xcode, Android SDK, Flutter and React Native tooling preinstalled",
        "You need SSH or VNC access to a running machine for debugging",
        "You need a choice of US or EU data centres for data residency",
        "You run GitHub Enterprise Server as well as GitHub Cloud"
      ],
      "cons": [
        "Pay-As-You-Go is off by default, so jobs do not run on Build Hub once the monthly budget is used up",
        "Pay-As-You-Go per-minute rates are twice the standard machine rates",
        "Personal access token authentication is available only on Enterprise plans",
        "M4 Pro, M5 Pro and M6 machines and Linux above 16 vCPU are listed only on the Enterprise plan"
      ],
      "verdict": "Bitrise Build Hub is a managed runner service for GitHub Actions from a mobile CI vendor, suited to teams that want Bitrise’s Mac hardware and mobile stacks without leaving GitHub Actions.",
      "use_cases": {
        "macos-ios": "Build Hub runs GitHub Actions jobs on Apple silicon Mac minis with Xcode stacks that Bitrise says are updated within 24 hours of a release.",
        "faster-builds": "Bitrise publishes its own benchmarks against GitHub-hosted runners and offers a co-located cache (vendor claims).",
        "security": "Bitrise states SOC 2 Type II and PCI DSS compliance and offers US and EU data centre locations."
      },
      "peers": [
        "github-hosted-runners",
        "getmac",
        "macstadium-orka",
        "cirrus-runners",
        "namespace",
        "warpbuild",
        "depot",
        "blacksmith"
      ],
      "sources": [
        "https://bitrise.io/platform/build-hub",
        "https://bitrise.io/pricing",
        "https://docs.bitrise.io/en/bitrise-build-hub",
        "https://docs.bitrise.io/en/bitrise-build-hub/build-hub-for-github-actions/build-hub-for-github-actions-overview",
        "https://docs.bitrise.io/en/bitrise-build-hub/build-hub-for-github-actions/creating-machine-pools-for-github-actions",
        "https://docs.bitrise.io/en/bitrise-build-hub/build-hub-for-github-actions/configuring-your-github-actions-workflow",
        "https://docs.bitrise.io/en/bitrise-build-hub/build-hub-for-github-actions/monitoring-machine-pools-for-github-actions",
        "https://docs.bitrise.io/en/bitrise-build-hub/infrastructure/build-machine-types",
        "https://docs.bitrise.io/en/bitrise-build-hub/infrastructure/ip-addresses-for-the-build-machines",
        "https://docs.bitrise.io/en/bitrise-build-hub/changelog",
        "https://bitrise.io/platform/devops/security",
        "https://bitrise.io/about",
        "https://bitrise.io/blog/post/m5-pro-runners-for-github-actions-coming-to-bitrise-build-hub",
        "https://bitrise.io/blog/post/ship-ios-and-android-builds-twice-as-fast-on-github-actions"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "flyci",
      "name": "FlyCI macOS Runners",
      "company": "FlyCI",
      "hq": null,
      "kind": "managed",
      "category": "macos",
      "website": "https://flyci.net",
      "docs": null,
      "pricing_url": null,
      "github": null,
      "tagline": "Managed Apple silicon macOS runners for GitHub Actions that FlyCI discontinued on September 30, 2024.",
      "ci_systems": [
        "GitHub Actions"
      ],
      "os": [
        "macOS"
      ],
      "arch": [
        "arm64"
      ],
      "isolation": null,
      "isolation_short": null,
      "isolation_class": null,
      "ephemeral": null,
      "open_source": null,
      "license": null,
      "self_host": false,
      "byoc": null,
      "clouds": null,
      "gpu": null,
      "gpu_detail": null,
      "hardware": null,
      "performance": null,
      "sizes": null,
      "max_size_short": null,
      "cache": null,
      "docker_layer_cache": null,
      "concurrency": null,
      "job_limits": null,
      "static_ips": null,
      "ssh_access": null,
      "observability": null,
      "setup": null,
      "runs_on": "flyci-macos-14-m2",
      "pricing": null,
      "price_short": null,
      "free_tier": null,
      "compliance": null,
      "regions": null,
      "best_for": "Historical reference only, since the runner service is no longer offered",
      "overview": [
        "**FlyCI macOS Runners** was a managed runner service for GitHub Actions. Teams installed the FlyCI GitHub App and pointed `runs-on` at a FlyCI label such as `flyci-macos-14-m2` to run jobs on FlyCI’s Apple silicon machines instead of GitHub-hosted macOS runners. In its own words, FlyCI started with Linux runners and later pivoted to macOS runners before adding an AI assistant named FlyCI Wingman.",
        "On August 14, 2024 FlyCI announced that it would discontinue the macOS runners effective September 30, 2024 and put its development effort into FlyCI Wingman, a tool that proposes fixes for failing GitHub Actions workflows. The announcement told customers to migrate, giving GitHub-hosted `macos-14` runners as an example. The FlyCI GitHub App listing now describes only Wingman. This record is kept so that older references to FlyCI runners can be resolved."
      ],
      "notable": [
        "August 14, 2024: FlyCI announced it would discontinue FlyCI macOS runners and focus on FlyCI Wingman",
        "September 30, 2024: effective date of the discontinuation given in the announcement"
      ],
      "pros": [],
      "cons": [
        "Discontinued: FlyCI’s announcement says the macOS runners are no longer available or supported after September 30, 2024"
      ],
      "verdict": "FlyCI macOS Runners was a managed macOS runner service for GitHub Actions that its operator discontinued on September 30, 2024.",
      "use_cases": {
        "macos-ios": "FlyCI offered Apple silicon macOS runners for GitHub Actions until it discontinued them on September 30, 2024."
      },
      "peers": [
        "github-hosted-runners",
        "getmac",
        "bitrise-runners",
        "cirrus-runners",
        "namespace",
        "warpbuild"
      ],
      "sources": [
        "https://web.archive.org/web/20251210141114/https://flyci.net/blog/flyci-discontinue-macos-runners",
        "https://web.archive.org/web/20240810080848/https://flyci.net/docs/getting-started/flyci-macos-runners",
        "https://github.com/apps/flyci-prod"
      ],
      "verified_on": "2026-11-22",
      "status": "discontinued",
      "status_note": "FlyCI announced on August 14, 2024 that it would discontinue FlyCI macOS runners effective September 30, 2024, after which they would no longer be available or supported, and that it would focus on FlyCI Wingman.",
      "alternatives": true
    },
    {
      "id": "veertu-anka",
      "name": "Anka Build Cloud",
      "company": "Veertu Inc.",
      "hq": {
        "city": "Los Altos",
        "region": "California",
        "country": "US",
        "source": "https://veertu.com/software-and-support-agreement/"
      },
      "kind": "byoc",
      "category": "macos",
      "website": "https://veertu.com/anka-build/",
      "docs": "https://docs.veertu.com/anka/",
      "pricing_url": null,
      "github": "https://github.com/veertuinc/anklet",
      "tagline": "Commercial macOS virtualization for CI: a controller and registry manage on-demand Anka VMs on your own Macs, hosted Macs or AWS EC2 Mac.",
      "ci_systems": [
        "GitHub Actions",
        "GitLab CI",
        "Buildkite",
        "Jenkins",
        "Azure Pipelines",
        "TeamCity"
      ],
      "os": [
        "macOS"
      ],
      "arch": [
        "arm64",
        "x64"
      ],
      "isolation": "Jobs run in Anka macOS virtual machines on Mac build nodes; the Anka GitHub Actions and the Anklet orchestrator start an on-demand VM per job and tear it down afterwards",
      "isolation_short": "macOS VM",
      "isolation_class": "vm",
      "ephemeral": true,
      "open_source": false,
      "license": null,
      "self_host": true,
      "byoc": true,
      "clouds": [
        "Bare metal",
        "AWS"
      ],
      "gpu": true,
      "gpu_detail": "Anka documents access to the Mac’s GPU inside VMs through the Metal API (paravirtualized graphics); no price is published.",
      "hardware": "Runs on Apple silicon and Intel Mac hosts that you supply, including AWS EC2 Mac instances.",
      "performance": null,
      "sizes": null,
      "max_size_short": null,
      "cache": "VM templates and tags are stored in the Anka Registry or an OCI registry, and nodes download only the deltas of a new template or tag.",
      "docker_layer_cache": null,
      "concurrency": "The Controller queues and load-balances VM requests across build nodes; licences are counted by host",
      "job_limits": null,
      "static_ips": null,
      "ssh_access": true,
      "observability": "The Controller dashboard shows build nodes, VM instances, templates and logs; Anklet can run an optional metrics aggregator.",
      "setup": "Install Anka on Mac build nodes and run the Controller and Registry (on macOS or as a Docker container), then connect CI through Anklet, the Anka GitHub Actions, the GitLab executor, or the Jenkins, TeamCity or Buildkite plugins.",
      "runs_on": "[self-hosted, \"${{ needs.action-up.outputs.action-id }}\"]",
      "pricing": "Prices are not published. Anka licences for Apple silicon are based on host count and host model, in Basic, Enterprise and Enterprise Plus tiers; Anka Develop is free for one VM at a time on Mac laptops.",
      "price_short": null,
      "free_tier": "30-day trial licence with unlimited access to product features; Anka Develop is free (one VM at a time, Mac laptops only)",
      "compliance": null,
      "regions": "Wherever your Mac hosts run: your own hardware, hosted Macs or AWS EC2 Mac",
      "best_for": "Organisations that run their own Mac fleet or EC2 Mac instances and want licensed macOS VM tooling with plugins for several CI systems",
      "overview": [
        "**Anka Build Cloud** is Veertu’s commercial software for running macOS VMs for CI. Anka Virtualization runs on each Mac build node; a Controller queues and schedules VM requests across nodes, and a Registry stores versioned VM templates and tags, in a workflow Veertu compares to Docker and Docker Hub. It runs on Apple silicon and Intel Macs that you own, rent from a host, or launch as AWS EC2 Mac instances.",
        "CI systems connect through plugins: GitHub Actions, a GitLab Runner custom executor, Jenkins, TeamCity, Buildkite and Azure DevOps Pipelines are listed for every Build licence tier. [Anklet](https://github.com/veertuinc/anklet), Veertu’s external orchestrator, listens for jobs and starts on-demand VMs without extra steps in workflow files. Licences are counted per host and host model in three tiers, and Veertu documents Metal GPU access inside VMs."
      ],
      "notable": [
        "Veertu’s docs list Anka Virtualization 3.10.2 and Anka Build Cloud 1.52.1 as current on 2026-11-22",
        "Anka 3.10.0 added thin OCI tags, which store a tool such as Xcode as its own tag that can be added to any VM",
        "Veertu says its earlier Veertu Desktop hypervisor was open-sourced in September 2016 and is no longer maintained"
      ],
      "pros": [
        "You operate your own Macs or EC2 Mac instances and want a controller, registry and CI plugins for them",
        "You use several CI systems: GitHub Actions, GitLab, Jenkins, TeamCity, Buildkite and Azure DevOps are listed for every Build tier",
        "You still build on Intel Macs as well as Apple silicon",
        "You need GPU access through Metal inside macOS VMs",
        "You want to keep VM images in an OCI registry you already use"
      ],
      "cons": [
        "Prices are not published on Veertu’s site",
        "SSO, permission groups and Controller API event logging are limited to the Enterprise Plus tier",
        "The Buildkite plugin does not use the Controller; it connects directly to a node agent",
        "State snapshot/suspend and USB device support are marked Intel only in the licence feature table"
      ],
      "verdict": "Anka Build Cloud is licensed macOS virtualization and orchestration software for teams that run CI on Mac hosts they control.",
      "use_cases": {
        "macos-ios": "Anka provisions on-demand macOS VMs for iOS and macOS CI on Apple silicon and Intel Mac hosts.",
        "byoc": "The software runs on your own Mac hardware or AWS EC2 Mac instances, with an official Anka AMI on AWS Marketplace.",
        "gitlab": "Veertu maintains a GitLab Runner custom executor that runs GitLab jobs in Anka VMs through the Controller."
      },
      "peers": [
        "macstadium-orka",
        "tart",
        "getmac",
        "bitrise-runners",
        "cirrus-runners",
        "github-hosted-runners",
        "gitlab-runner"
      ],
      "sources": [
        "https://veertu.com/anka-build/",
        "https://veertu.com/about-contact/",
        "https://veertu.com/software-and-support-agreement/",
        "https://docs.veertu.com/anka/",
        "https://docs.veertu.com/anka/licensing/",
        "https://docs.veertu.com/anka/anka-build-cloud/build-license-tiers/",
        "https://docs.veertu.com/anka/plugins-and-integrations/controller-+-registry/github-actions/",
        "https://docs.veertu.com/anka/plugins-and-integrations/controller-+-registry/gitlab-custom-executor/",
        "https://docs.veertu.com/anka/plugins-and-integrations/controller-less/buildkite/",
        "https://github.com/veertuinc/anklet"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "cilicon",
      "name": "Cilicon",
      "company": "Trade Republic",
      "hq": null,
      "kind": "open-source",
      "category": "macos",
      "website": "https://github.com/traderepublic/Cilicon",
      "docs": "https://github.com/traderepublic/Cilicon#readme",
      "pricing_url": null,
      "github": "https://github.com/traderepublic/Cilicon",
      "tagline": "Open-source macOS app that runs ephemeral CI virtual machines on a single Apple silicon Mac using Apple’s Virtualization framework.",
      "ci_systems": [
        "GitHub Actions",
        "GitLab CI",
        "Buildkite"
      ],
      "os": [
        "macOS"
      ],
      "arch": [
        "arm64"
      ],
      "isolation": "Each job runs in an ephemeral macOS virtual machine created with Apple’s Virtualization framework and provisioned over SSH",
      "isolation_short": "macOS VM",
      "isolation_class": "vm",
      "ephemeral": true,
      "open_source": true,
      "license": "MIT",
      "self_host": true,
      "byoc": true,
      "clouds": [
        "Bare metal"
      ],
      "gpu": null,
      "gpu_detail": null,
      "hardware": "An Apple silicon Mac that you supply; the README recommends a machine fully dedicated to the task.",
      "performance": "The README describes VM performance as near-native (project claim)",
      "sizes": null,
      "max_size_short": null,
      "cache": null,
      "docker_layer_cache": null,
      "concurrency": "One VM at a time per host; running two VMs in parallel is listed under ideas for the future",
      "job_limits": null,
      "static_ips": null,
      "ssh_access": null,
      "observability": null,
      "setup": "Download the app, write a cilicon.yml in the host’s home directory with an image source and a provisioner (GitHub App, GitLab runner token or Buildkite agent token), then add Cilicon as a launch item and enable automatic login.",
      "runs_on": null,
      "pricing": "Free and MIT-licensed; you supply the Mac.",
      "price_short": null,
      "free_tier": null,
      "compliance": null,
      "regions": "Wherever your Mac runs",
      "best_for": "Teams with one or a few dedicated Apple silicon Macs that want ephemeral self-hosted runners without an orchestration layer",
      "overview": [
        "**Cilicon** is a macOS app from Trade Republic that turns an Apple silicon Mac into a self-hosted CI machine. It uses Apple’s Virtualization framework to create a VM from an image, connects over SSH, starts a runner, and discards the VM when the job ends, then repeats the cycle. Provisioners are included for GitHub Actions, GitLab Runner, Buildkite Agent and arbitrary scripts.",
        "Configuration is a single `cilicon.yml` file naming an image source and a provisioner. Since version 2.0 Cilicon has used the Tart image format and ships an OCI client, so it can download prebuilt macOS and Xcode images from a registry. The README recommends a freshly restored, dedicated Mac with automatic login, and says Trade Republic itself uses GitHub Actions for its iOS builds."
      ],
      "notable": [
        "December 12, 2025: release v2.4.2 published on GitHub",
        "Version 2.0 added a built-in SSH client and partial adoption of the Tart image format",
        "The repository’s most recent push recorded by GitHub is dated September 14, 2026"
      ],
      "pros": [
        "You have a spare Apple silicon Mac and want ephemeral runners on it with one app and one config file",
        "You want MIT-licensed tooling with no licence fee",
        "You use GitHub Actions, GitLab or Buildkite, each of which has a built-in provisioner",
        "You want to reuse prebuilt macOS and Xcode images published in the Tart format"
      ],
      "cons": [
        "Running two VMs in parallel is listed as a future idea, so a host runs one job at a time",
        "Logging and monitoring are listed as ideas for the future",
        "The liveness probe is supported only for the GitHub Actions provisioner",
        "The README warns of an SSH issue on macOS 15.0 to 15.3 and recommends macOS 15.4 or later"
      ],
      "verdict": "Cilicon is a small MIT-licensed macOS app for teams that want ephemeral self-hosted CI VMs on individual Apple silicon Macs.",
      "use_cases": {
        "macos-ios": "Cilicon runs each CI job in a fresh macOS VM on Apple silicon, using images with Xcode preinstalled.",
        "open-source": "The app is MIT-licensed and runs entirely on a Mac you operate.",
        "gitlab": "A GitLab Runner provisioner registers a runner inside each VM using an authentication token.",
        "byoc": "Everything runs on your own Mac hardware."
      },
      "peers": [
        "tart",
        "macstadium-orka",
        "getmac",
        "actions-runner-controller",
        "gitlab-runner",
        "github-hosted-runners"
      ],
      "sources": [
        "https://github.com/traderepublic/Cilicon",
        "https://github.com/traderepublic/Cilicon/blob/main/README.md",
        "https://github.com/traderepublic/Cilicon/releases"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "tartelet",
      "name": "Tartelet",
      "company": "Framna Denmark",
      "hq": null,
      "kind": "open-source",
      "category": "macos",
      "website": "https://github.com/framna-dk/tartelet",
      "docs": "https://github.com/framna-dk/tartelet/wiki",
      "pricing_url": null,
      "github": "https://github.com/framna-dk/tartelet",
      "tagline": "Open-source macOS app that runs up to two GitHub Actions runners in ephemeral Tart virtual machines on a single Mac.",
      "ci_systems": [
        "GitHub Actions"
      ],
      "os": [
        "macOS"
      ],
      "arch": [
        "arm64"
      ],
      "isolation": "Each runner runs in a Tart virtual machine that is cloned before a job and shut down and deleted after it",
      "isolation_short": "Tart VM",
      "isolation_class": "vm",
      "ephemeral": true,
      "open_source": true,
      "license": "MIT",
      "self_host": true,
      "byoc": true,
      "clouds": [
        "Bare metal"
      ],
      "gpu": null,
      "gpu_detail": null,
      "hardware": "An Apple silicon Mac that you supply; the maintainers’ own figures come from a 2020 Mac mini M1 with 16 GB of memory.",
      "performance": "The maintainers report jobs running 3 to 4 times faster on a Mac mini M1 than on GitHub’s runners, measured in January/February 2023 (project claim)",
      "sizes": null,
      "max_size_short": null,
      "cache": null,
      "docker_layer_cache": null,
      "concurrency": "Up to two virtual machines per host, which the README gives as the maximum Apple’s Virtualization framework allows",
      "job_limits": null,
      "static_ips": null,
      "ssh_access": null,
      "observability": null,
      "setup": "Install Tart and Tartelet on the host Mac, create a GitHub App with self-hosted runner permissions, build a Tart VM image, and select both in Tartelet’s settings.",
      "runs_on": null,
      "pricing": "Free and MIT-licensed; you supply the Mac.",
      "price_short": null,
      "free_tier": null,
      "compliance": null,
      "regions": "Wherever your Mac runs",
      "best_for": "Small teams on GitHub Actions that want two isolated, ephemeral runners per Mac managed from a desktop app",
      "overview": [
        "**Tartelet** is a macOS app that launches self-hosted GitHub Actions runners inside virtual machines managed by [Tart](https://tart.run). For each runner it clones a VM, boots it, runs a setup script that downloads GitHub’s runner application and registers it, and waits for a job. When the job finishes the runner removes itself, the VM is shut down and deleted, and the cycle starts again.",
        "The app manages up to two VMs on one host, which its README gives as the limit of Apple’s Virtualization framework, and reports that recreating a VM takes about 25 to 30 seconds. Runners register through a GitHub App at organisation or repository scope, with the private key kept in the host’s keychain. The project is MIT-licensed and now lives under the Framna Denmark GitHub organisation, having started at shapehq/tartelet."
      ],
      "notable": [
        "June 13, 2025: release 0.12.0 published on GitHub",
        "The repository’s most recent push recorded by GitHub is dated February 10, 2026",
        "The README’s performance figures were last updated in January/February 2023"
      ],
      "pros": [
        "You want two parallel GitHub Actions runners on one Mac, each job in a VM that is recreated afterwards",
        "You prefer a desktop app with settings panes to writing orchestration config",
        "You want MIT-licensed tooling with no licence fee",
        "You already build VM images with Tart"
      ],
      "cons": [
        "Supports GitHub Actions only",
        "At most two virtual machines per host",
        "Requires Tart to be installed separately on the host",
        "Running two VMs in parallel made jobs about 12% slower than one VM in the maintainers’ tests"
      ],
      "verdict": "Tartelet is an MIT-licensed desktop app for teams that want a pair of ephemeral GitHub Actions runners on each Apple silicon Mac they own.",
      "use_cases": {
        "macos-ios": "Tartelet runs GitHub Actions jobs in fresh macOS VMs on a Mac you own.",
        "open-source": "The app is MIT-licensed and self-hosted on your own hardware.",
        "byoc": "Runners live on your own Mac; nothing is hosted by a vendor."
      },
      "peers": [
        "tart",
        "macstadium-orka",
        "getmac",
        "actions-runner-controller",
        "cirrus-runners",
        "github-hosted-runners"
      ],
      "sources": [
        "https://github.com/framna-dk/tartelet",
        "https://github.com/framna-dk/tartelet/blob/main/README.md",
        "https://github.com/framna-dk/tartelet/wiki/Configuring-Tartelet",
        "https://github.com/framna-dk/tartelet/wiki/Installing-Tartelet",
        "https://github.com/framna-dk/tartelet/releases"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "blacksmith",
      "name": "Blacksmith",
      "company": "Blacksmith Software Inc.",
      "hq": {
        "city": "San Francisco",
        "region": "California",
        "country": "US",
        "source": "https://docs.blacksmith.sh/about/privacy-policy"
      },
      "kind": "managed",
      "category": "managed-github-actions",
      "website": "https://www.blacksmith.sh",
      "docs": "https://docs.blacksmith.sh",
      "pricing_url": "https://www.blacksmith.sh/pricing",
      "github": null,
      "tagline": "Managed GitHub Actions runners in Firecracker microVMs on the vendor’s own bare-metal fleet, with caches kept next to the runners.",
      "ci_systems": [
        "GitHub Actions"
      ],
      "os": [
        "Linux",
        "macOS",
        "Windows"
      ],
      "arch": [
        "x64",
        "arm64"
      ],
      "isolation": "Each Linux and Windows job runs in an ephemeral Firecracker microVM on Blacksmith’s bare-metal hosts and is destroyed afterwards; the isolation used for macOS runners is not documented",
      "isolation_short": "Firecracker microVM",
      "isolation_class": "microvm",
      "ephemeral": true,
      "open_source": false,
      "license": null,
      "self_host": false,
      "byoc": null,
      "clouds": null,
      "gpu": null,
      "gpu_detail": null,
      "hardware": "Blacksmith describes its hosts as bare-metal machines with gaming CPUs chosen for single-core performance and NVMe drives, in ISO 27001 data centers; CPU models are not named, and macOS runners use Apple M4.",
      "performance": "Blacksmith says most CI jobs run about 2x faster than on GitHub-hosted runners and cache downloads about 4x faster, and that customers report 2x to 40x faster Docker builds (vendor claims)",
      "sizes": "2 to 32 vCPU with 8 to 128 GB RAM on Linux x64; 2 to 32 vCPU with 6 to 96 GB on Linux arm64; 2 to 32 vCPU with 7 to 112 GB on Windows x64 (public beta); 6 or 12 vCPU with 24 or 48 GB on macOS",
      "max_size_short": "32 vCPU / 128 GB",
      "cache": "Standard cache actions (actions/cache and the setup-* actions) are served from a cache in the same data center as the runners, with 25 GB of free storage per repository per week and eviction of entries unused for 7 days. Docker layers persist on NVMe-backed sticky disks through useblacksmith/setup-docker-builder, billed at $0.50/GB/month; the docs also cover Git checkout, container image and Bazel, Go, Turborepo and Nx build caching.",
      "docker_layer_cache": true,
      "concurrency": "Blacksmith documents no limit on concurrent jobs or vCPUs",
      "job_limits": null,
      "static_ips": true,
      "ssh_access": true,
      "observability": "Console with run history, searchable logs, metrics, monitors, test analytics and CI analytics.",
      "setup": "Install the Blacksmith GitHub App on a GitHub organization, then change runs-on in each workflow by hand or with the Migration Wizard. Personal repositories are not supported.",
      "runs_on": "blacksmith-4vcpu-ubuntu-2404",
      "pricing": "The pricing page lists Ubuntu x64 at $0.004/min, Ubuntu ARM at $0.0025/min, Windows x64 at $0.008/min and macOS M4 at $0.08/min; the docs say runners cost half of GitHub’s per-minute rate. Add-ons: Docker layer caching and sticky disks at $0.50/GB/month and static IPs at $100/IP/month. Flex scheduling costs 40% less for jobs that can wait up to 24 hours to start.",
      "price_short": "$0.004/min",
      "free_tier": "3,000 free minutes per organization per month, counted as x64 2-vCPU minutes; larger, ARM, Windows and macOS runners consume them at weighted rates",
      "compliance": "SOC 2 Type 2; GDPR compliant (as stated by Blacksmith); HIPAA described as in progress",
      "regions": "us-east (Ashburn), us-west (Phoenix), eu-west (Amsterdam) and eu-central (Frankfurt); changing region requires contacting support",
      "best_for": "Teams on GitHub Actions that want microVM-isolated runners on bare metal with built-in Docker layer caching and CI observability",
      "overview": [
        "**Blacksmith** is a managed replacement for GitHub-hosted runners. After installing its GitHub App on an organization, a team changes the `runs-on` label and jobs run on Blacksmith’s own fleet of bare-metal machines in the US and EU. Linux and Windows jobs each get a Firecracker microVM that is destroyed when the job ends. Runner images follow GitHub’s official images, and sizes range from 2 to 32 vCPU on Linux and Windows, with 6 and 12 vCPU Apple M4 macOS runners.",
        "The service pairs compute with caching and tooling: standard cache actions are redirected to a co-located cache, Docker layers persist on NVMe-backed sticky disks, and a console provides logs, metrics, test analytics and SSH access to running jobs. Blacksmith documents no concurrency limits and offers [four regions](https://docs.blacksmith.sh/blacksmith-runners/overview). The fleet is multi-tenant on shared hardware. It is available to GitHub organizations only, with GitHub Enterprise Server and data-residency setups arranged through support."
      ],
      "notable": [
        "Blacksmith announced a Series C led by Sequoia Capital at a $1B valuation on 21 October 2026, bringing total funding to $125M.",
        "A $45M Series B led by Peak XV Partners was announced on 12 August 2026; Blacksmith says the round closed in March 2026.",
        "Blacksmith published a post on building its macOS runners on 16 April 2026."
      ],
      "pros": [
        "You want each Linux or Windows job in its own Firecracker microVM that is destroyed after the job",
        "You want Docker layers kept on persistent NVMe-backed disks between runs",
        "You need to run many jobs at once: Blacksmith documents no concurrency limit",
        "You want to SSH into running jobs and search CI logs from one console",
        "You need Linux, Windows and macOS runners from one vendor, with a 3,000-minute monthly free allowance"
      ],
      "cons": [
        "Available to GitHub organizations only, not personal repositories",
        "Windows runners are in public beta and do not support Docker Linux containers",
        "KVM and nested virtualization are supported on x64 Linux only, not ARM",
        "Docker layer caching and sticky disks are billed separately at $0.50/GB/month"
      ],
      "verdict": "A managed GitHub Actions runner service on vendor-owned bare metal with microVM isolation, suited to organizations that want faster hardware plus integrated caching and observability.",
      "use_cases": {
        "faster-builds": "Blacksmith runs jobs on bare-metal gaming CPUs with a co-located cache and says most CI jobs run about 2x faster (vendor claim).",
        "lower-cost": "Listed rates start at $0.004/min for Ubuntu x64, which Blacksmith says is half of GitHub’s per-minute rate, with 3,000 free minutes per month.",
        "docker-builds": "Docker layers persist across runs on NVMe-backed sticky disks through Blacksmith’s Docker builder actions.",
        "arm64": "Native arm64 Linux runners from 2 to 32 vCPU are listed at $0.0025/min.",
        "macos-ios": "Apple M4 macOS runners with 6 or 12 vCPU are available with macOS 15, 26 and 27 images.",
        "windows": "Windows Server 2025 x64 runners from 2 to 32 vCPU are offered in public beta."
      },
      "peers": [
        "depot",
        "namespace",
        "warpbuild",
        "ubicloud",
        "luxor-runners",
        "github-hosted-runners"
      ],
      "sources": [
        "https://www.blacksmith.sh/",
        "https://www.blacksmith.sh/pricing",
        "https://www.blacksmith.sh/security",
        "https://docs.blacksmith.sh/blacksmith-runners/overview",
        "https://docs.blacksmith.sh/blacksmith-runners/concurrency",
        "https://docs.blacksmith.sh/blacksmith-runners/static-ip",
        "https://docs.blacksmith.sh/blacksmith-caching/dependencies-actions",
        "https://docs.blacksmith.sh/blacksmith-caching/docker-builds",
        "https://docs.blacksmith.sh/blacksmith-observability/ssh-access",
        "https://docs.blacksmith.sh/introduction/quickstart",
        "https://docs.blacksmith.sh/introduction/github-enterprise",
        "https://docs.blacksmith.sh/about/privacy-policy",
        "https://www.blacksmith.sh/blog/series-c",
        "https://www.blacksmith.sh/blog/announcing-blacksmiths-series-b-led-by-peak-xv-partners",
        "https://www.blacksmith.sh/blog"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "depot",
      "name": "Depot GitHub Actions Runners",
      "company": "Depot Technologies Inc.",
      "hq": null,
      "kind": "managed",
      "category": "managed-github-actions",
      "website": "https://depot.dev",
      "docs": "https://depot.dev/docs/github-actions/overview",
      "pricing_url": "https://depot.dev/pricing",
      "github": "https://github.com/depot/cli",
      "tagline": "Managed GitHub Actions runners on single-tenant AWS EC2 instances, connected to Depot’s cache and container build service.",
      "ci_systems": [
        "GitHub Actions"
      ],
      "os": [
        "Linux",
        "macOS",
        "Windows"
      ],
      "arch": [
        "x64",
        "arm64"
      ],
      "isolation": "Each job runs on a new single-tenant EC2 instance that has never been used before and is destroyed when the job completes",
      "isolation_short": "Single-tenant EC2 instance",
      "isolation_class": "vm",
      "ephemeral": true,
      "open_source": false,
      "license": null,
      "self_host": false,
      "byoc": true,
      "clouds": [
        "AWS"
      ],
      "gpu": true,
      "gpu_detail": "Depot’s docs list GPU-enabled runners as available on the Business plan only; models and prices are not published.",
      "hardware": "AWS EC2 instances: 4th Gen AMD EPYC Genoa for x64 Linux, Graviton4 for arm64, Intel for Windows, and Apple M2 (macOS 14 and 15) or M4 (macOS 26) for macOS; part of the memory is used as a RAM disk on Linux and macOS.",
      "performance": "Depot says its runners are up to 3x faster than GitHub-hosted runners, with 10x faster caching (vendor claim)",
      "sizes": "2 to 64 vCPU with 8 to 256 GB RAM on Linux x64, Linux arm64 and Windows; macOS runners are 8 vCPU with 24 GB",
      "max_size_short": "64 vCPU / 256 GB",
      "cache": "Runners are preconfigured to use Depot Cache as the GitHub Actions cache backend (14-day retention and no size limit by default); it also serves Bazel, Go, Gradle, Turborepo, Nx, sccache and Pants. Docker layer caching comes from Depot’s separately metered container build service, which keeps a persistent layer cache on NVMe SSDs for each project.",
      "docker_layer_cache": true,
      "concurrency": "Unlimited concurrency on all plans; macOS runners come from a fixed pool, so jobs can queue when demand spikes",
      "job_limits": null,
      "static_ips": true,
      "ssh_access": null,
      "observability": "GitHub Actions analytics with CPU and memory utilisation, runner sizing recommendations, step-level timing, failure rates and live runner status.",
      "setup": "Connect a GitHub organization by installing the Depot GitHub App from the Depot dashboard (organization owner role required), then change runs-on to a Depot label. Repositories owned by personal accounts are not supported.",
      "runs_on": "depot-ubuntu-24.04",
      "pricing": "Plans: Developer at $20/month with 2,000 GitHub Actions minutes and 25 GB of cache, Startup at $200/month with 20,000 minutes and 250 GB, Business custom. Additional minutes cost $0.006/min for the 2-vCPU Linux runner (x64 or arm64), scaling with size to $0.192/min at 64 vCPU; the runner-types page lists Windows from $0.008/min and macOS at $0.08/min. Usage is tracked per second with no one-minute minimum, and extra cache storage is $0.20/GB/month.",
      "price_short": "$0.006/min",
      "free_tier": "7-day free trial, no credit card required",
      "compliance": "SOC 2 Type 2",
      "regions": "Depot says it runs on AWS in the United States, and in the European Union when a project selects the EU region; the runner docs do not name specific regions",
      "best_for": "Teams on GitHub Actions that want single-tenant EC2 runners alongside a shared build cache and remote container builds from the same vendor",
      "overview": [
        "**Depot GitHub Actions Runners** are managed runners that replace GitHub-hosted ones through a `runs-on` label such as `depot-ubuntu-24.04`. GitHub still orchestrates the workflow; Depot receives the job webhook and assigns a fresh EC2 instance from a standby pool, then terminates it when the job completes. Linux runners use AMD EPYC Genoa or Graviton4 processors in sizes from 2 to 64 vCPU, and Windows and macOS runners are also offered.",
        "Runners are one part of a wider platform: they are preconfigured for [Depot Cache](https://depot.dev/docs/cache/overview) and sit in the same private network as Depot’s remote container builders, which keep Docker layers on persistent NVMe storage. Billing is by plan with included minutes, tracked per second. Egress filtering is available on all plans, while static outbound IPs, GPU runners, custom images and deployment in your own AWS account are Business-plan features. Only repositories owned by GitHub organizations are supported."
      ],
      "notable": [
        "Depot was incorporated in January 2022 and launched publicly in July 2022, according to its about page.",
        "Depot completed a $4.1M seed round in August 2024.",
        "Depot raised a $10M Series A in March 2026."
      ],
      "pros": [
        "You want every job on a fresh single-tenant EC2 instance that is destroyed afterwards",
        "You want per-second usage tracking with no one-minute minimum",
        "You need large Linux runners: up to 64 vCPU and 256 GB RAM on x64 and arm64",
        "You already use, or plan to use, Depot’s container builds and remote cache with the same account",
        "You may later need the runners in your own AWS account, which Depot offers on its Business plan"
      ],
      "cons": [
        "Plans start at $20/month after a 7-day trial; a permanent free allowance is not documented",
        "Only repositories owned by GitHub organizations are supported",
        "macOS capacity is a fixed pool, so macOS jobs can queue when demand spikes",
        "Windows runners have no Hyper-V, so Docker-dependent workloads are unlikely to work"
      ],
      "verdict": "A managed GitHub Actions runner service built on single-tenant AWS instances, suited to teams that also want Depot’s cache and container build service.",
      "use_cases": {
        "faster-builds": "Depot says its runners are up to 3x faster than GitHub-hosted runners and come preconfigured for its distributed cache (vendor claim).",
        "docker-builds": "Depot’s container build service keeps a persistent Docker layer cache on NVMe SSDs and is used from Actions through depot/build-push-action.",
        "arm64": "Graviton4 arm64 runners from 2 to 64 vCPU are priced the same as x64.",
        "byoc": "Depot’s docs say runners can run in its cloud or in your AWS account, the latter on the Business plan.",
        "security": "Each job gets a never-reused single-tenant EC2 instance, with egress filtering on all plans and SOC 2 Type 2 compliance stated."
      },
      "peers": [
        "blacksmith",
        "namespace",
        "warpbuild",
        "ubicloud",
        "runs-on",
        "github-hosted-runners"
      ],
      "sources": [
        "https://depot.dev/pricing",
        "https://depot.dev/docs/github-actions/overview",
        "https://depot.dev/docs/github-actions/runner-types",
        "https://depot.dev/docs/github-actions/quickstart",
        "https://depot.dev/docs/github-actions/troubleshooting",
        "https://depot.dev/products/github-actions",
        "https://depot.dev/docs/security",
        "https://depot.dev/security",
        "https://depot.dev/docs/cache/overview",
        "https://depot.dev/docs/container-builds/overview",
        "https://depot.dev/docs/managed/overview",
        "https://depot.dev/about",
        "https://depot.dev/terms-of-service"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "namespace",
      "name": "Namespace",
      "company": "Namespace Labs Inc.",
      "hq": null,
      "kind": "managed",
      "category": "managed-github-actions",
      "website": "https://namespace.so",
      "docs": "https://namespace.so/docs/solutions/github-actions",
      "pricing_url": "https://namespace.so/pricing",
      "github": null,
      "tagline": "Managed runners for GitHub Actions and Buildkite on the vendor’s own hardware, configured through profiles with persistent cache volumes.",
      "ci_systems": [
        "GitHub Actions",
        "Buildkite"
      ],
      "os": [
        "Linux",
        "macOS",
        "Windows"
      ],
      "arch": [
        "x64",
        "arm64"
      ],
      "isolation": "Every job runs in its own instance inside a virtual machine that is discarded when the job ends; the hypervisor is not named",
      "isolation_short": "Virtual machine",
      "isolation_class": "vm",
      "ephemeral": true,
      "open_source": false,
      "license": null,
      "self_host": false,
      "byoc": null,
      "clouds": null,
      "gpu": null,
      "gpu_detail": null,
      "hardware": "Namespace says it deploys its own hardware in its own racks: AMD EPYC for Linux and Windows amd64, AmpereOne for Linux arm64 (Linux on Apple Silicon is in early access), and Apple M4 Pro and M5 Max for macOS.",
      "performance": null,
      "sizes": "Linux up to 64 vCPU and 512 GB RAM, Windows up to 32 vCPU and 512 GB, macOS up to 16 vCPU and 56 GB; standard label shapes run from 2 vCPU / 4 GB to 32 vCPU / 64 GB, and instances over 64 GB RAM need special access",
      "max_size_short": "64 vCPU / 512 GB",
      "cache": "Cache Volumes persist data across runs; a runner profile can cache container images, toolchain downloads, action downloads and Git checkouts, and nscloud-cache-action replaces actions/cache. Docker builds go to Remote Builders with layer caching by default, and there are cache integrations for Bazel, Turborepo, Pants and Moonrepo.",
      "docker_layer_cache": true,
      "concurrency": "Per plan: Developer 32 Linux vCPU and 12 macOS vCPU; Team 64 Linux, 24 macOS and 32 Windows vCPU; Business 160 Linux, 48 macOS and 80 Windows vCPU; custom on Enterprise",
      "job_limits": "Maximum instance duration is 3 hours on Developer, 5 hours on Team and up to 24 hours on Business and Enterprise",
      "static_ips": true,
      "ssh_access": true,
      "observability": "Per-job CPU, memory, disk and network metrics correlated with steps, container and build logs, out-of-memory detection, crash dumps and queue-time insights.",
      "setup": "Connect a GitHub organization or personal account in the Namespace dashboard, create a runner profile (OS, architecture, shape and caching), then set runs-on to the profile label.",
      "runs_on": "nscloud-ubuntu-22.04-amd64-4x8",
      "pricing": "Billed in unit minutes (vCPU × minutes × platform multiplier: Linux 1, Windows 2, macOS 10), with a 1-minute minimum. Team is $100/month with 100,000 unit minutes and Business $250/month with 250,000; Developer is pay as you go. A 2 vCPU / 4 GB Linux runner is listed at $0.002/min prepaid and $0.003/min overage, 2 vCPU / 8 GB at $0.004 and $0.006, and macOS 6 vCPU / 14 GB at $0.06 and $0.09. Cache volume storage is $0.0048/GB-day on Developer, with free allowances on Team and Business.",
      "price_short": "$0.002/min prepaid, $0.003/min overage (2 vCPU / 4 GB)",
      "free_tier": "30-day free trial of Developer plan features, no credit card required",
      "compliance": "SOC 2 Type II",
      "regions": "Specific regions are not named; Namespace says it runs its own hardware at multiple sites, offers a regional preference on all plans and region-exclusive residency on enterprise plans",
      "best_for": "Teams on GitHub Actions or Buildkite that want VM-isolated Linux, macOS and Windows runners with persistent cache volumes and fine-grained machine shapes",
      "overview": [
        "**Namespace** runs CI jobs on a compute platform it operates on its own hardware. For GitHub Actions, a team connects its organization, defines runner profiles that fix the OS, architecture, machine shape and caching, and points `runs-on` at a profile label. Namespace says every job runs in its own virtual machine that is discarded when the job ends. The same platform provides ephemeral Buildkite agents, remote Docker builders, Bazel remote execution and Devboxes.",
        "Caching is built around Cache Volumes that persist between runs and are shared by all jobs using a profile, with optional protection so only chosen branches can write. Usage is billed in unit minutes that scale with vCPU count and a platform multiplier, and plans set concurrency and instance-duration limits. Debugging options include a dashboard terminal, a breakpoint action and VNC or RDP sessions. [Egress policies](https://namespace.so/docs/security/egress-policy) and stable egress IP ranges are documented; dedicated egress pools are for enterprise customers."
      ],
      "notable": [
        "Namespace announced a $42M Series B led by Scale Venture Partners on 5 October 2026, bringing total funding to $65M.",
        "The Series B post says the Series A, led by NEA, was raised seven months earlier.",
        "Git Snapshots for GitHub Actions and Buildkite runners were announced on 30 September 2026.",
        "A post dated 5 May 2026 describes Namespace’s security model: one VM per job, discarded when the job ends."
      ],
      "pros": [
        "You want each job in its own virtual machine that is discarded when the job ends",
        "You want cache volumes that persist container images, toolchains, action downloads and Git checkouts between runs",
        "You need Linux, macOS and Windows runners, with Linux shapes up to 64 vCPU and 512 GB RAM",
        "You run Buildkite as well as GitHub Actions and want one provider for both",
        "You need stable egress IP ranges and egress policies for allowlisting and network control"
      ],
      "cons": [
        "Concurrency is capped per plan, for example 32 Linux vCPU on Developer and 64 on Team",
        "Maximum instance duration is 3 hours on Developer and 5 hours on Team",
        "Windows runners are not available on the Developer plan",
        "Native SSH to instances is not enabled by default and requires enrolment through support"
      ],
      "verdict": "A managed runner platform for GitHub Actions and Buildkite on vendor-operated hardware, suited to teams that want per-job VMs, persistent cache volumes and all three operating systems.",
      "use_cases": {
        "faster-builds": "Cache Volumes keep container images, toolchains, action downloads and Git checkouts between runs, and Docker builds use Remote Builders with layer caching.",
        "docker-builds": "Docker builds on Namespace runners are sent to Remote Builders by default, with layer caching included.",
        "arm64": "Linux arm64 runners use AmpereOne processors, with Linux on Apple Silicon in early access.",
        "macos-ios": "macOS runners use Apple M4 Pro and M5 Max hardware with up to 16 vCPU and 56 GB RAM, with VNC remote display for debugging.",
        "windows": "Windows amd64 runners on AMD EPYC are included in the Team and Business plans.",
        "security": "Each job runs in its own discarded VM, and egress policies can restrict outbound traffic per runner profile; SOC 2 Type II is stated."
      },
      "peers": [
        "blacksmith",
        "depot",
        "warpbuild",
        "ubicloud",
        "cirrus-runners",
        "buildkite-hosted-agents",
        "github-hosted-runners"
      ],
      "sources": [
        "https://namespace.so/",
        "https://namespace.so/pricing",
        "https://namespace.so/docs/solutions/github-actions",
        "https://namespace.so/docs/reference/github-actions/runner-configuration",
        "https://namespace.so/docs/solutions/github-actions/docker-builds",
        "https://namespace.so/docs/solutions/github-actions/debugging/interactive-access",
        "https://namespace.so/docs/architecture/compute",
        "https://namespace.so/docs/architecture/compute/resource-limits",
        "https://namespace.so/docs/architecture/networking/security",
        "https://namespace.so/docs/security/egress-policy",
        "https://namespace.so/docs/workspaces/data-residency",
        "https://namespace.so/docs/workspaces/security",
        "https://namespace.so/docs/solutions/buildkite-agents",
        "https://namespace.so/blog/defense-in-depth-how-we-think-about-security",
        "https://namespace.so/blog/series-b",
        "https://namespace.so/blog",
        "https://namespace.so/terms"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "warpbuild",
      "name": "WarpBuild",
      "company": "Argonaut HQ, Inc",
      "hq": null,
      "kind": "managed",
      "category": "managed-github-actions",
      "website": "https://www.warpbuild.com",
      "docs": "https://www.warpbuild.com/docs/ci",
      "pricing_url": "https://www.warpbuild.com/pricing",
      "github": "https://github.com/WarpBuilds/warpbuild-agent",
      "tagline": "Managed GitHub Actions runners in ephemeral VMs, also offered as a deployment in your own AWS, Google Cloud or Azure account.",
      "ci_systems": [
        "GitHub Actions"
      ],
      "os": [
        "Linux",
        "macOS",
        "Windows"
      ],
      "arch": [
        "x64",
        "arm64"
      ],
      "isolation": "Each runner runs in its own virtual machine with an encrypted storage volume, created on demand and destroyed after each build",
      "isolation_short": "Ephemeral VM",
      "isolation_class": "vm",
      "ephemeral": true,
      "open_source": false,
      "license": null,
      "self_host": false,
      "byoc": true,
      "clouds": [
        "AWS",
        "Google Cloud",
        "Azure"
      ],
      "gpu": null,
      "gpu_detail": null,
      "hardware": "The cloud provider and CPU models behind the standard cloud runners are not documented; macOS runners and the “xfast” Linux arm64 runners use Apple M4 Pro and M5 Pro hardware.",
      "performance": "WarpBuild advertises runners that are “2x faster, 50% cheaper” and 3-10x faster caching (vendor claims)",
      "sizes": "2 to 32 vCPU with 8 to 128 GB RAM on Linux x64 and arm64; 4 to 32 vCPU with 16 to 128 GB on Windows; 6 or 12 vCPU with 14 or 28 GB on macOS",
      "max_size_short": "32 vCPU / 128 GB",
      "cache": "WarpBuild Cache (WarpBuilds/cache) is a drop-in replacement for actions/cache, enabled by default on Linux runners, with entries expiring 7 days after last use. Docker layer caching is offered through remote Docker Builders or a type=gha cache endpoint, and snapshot runners can save and boot from full VM disk snapshots on cloud Ubuntu runners.",
      "docker_layer_cache": true,
      "concurrency": "Unlimited concurrency on Linux and Windows for generally available features; high-concurrency macOS requires contacting support",
      "job_limits": null,
      "static_ips": true,
      "ssh_access": true,
      "observability": "CPU, memory, filesystem and network metrics, system and GitHub Actions logs, right-sizing recommendations, and export to your own OTLP backend.",
      "setup": "Sign up, install the WarpBuild GitHub App on a GitHub organization, then change runs-on to a WarpBuild label; an MCP-based flow can open the pull request for you. For BYOC, connect a cloud account and create a stack first. Personal accounts are not supported.",
      "runs_on": "warp-ubuntu-latest-x64-4x",
      "pricing": "Cloud runners per minute: Linux x64 from $0.004 (2 vCPU) to $0.064 (32 vCPU); Linux arm64 from $0.003 to $0.048; Windows from $0.016 (4 vCPU) to $0.128; macOS $0.08 (6 vCPU) and $0.16 (12 vCPU). Cache storage is $0.20 per GB-month plus $0.0001 per operation, and snapshot restore is $0.04 per job. BYOC runners carry a $0.002/min fee, with infrastructure billed by your cloud provider and cache and snapshots free.",
      "price_short": "$0.004/min",
      "free_tier": null,
      "compliance": "SOC 2 Type 2 (Security, Availability and Confidentiality criteria)",
      "regions": "Not listed for the standard cloud runners; region-specific infrastructure is offered on enterprise plans, and BYOC runners run in the region of your own stack",
      "best_for": "Teams on GitHub Actions that want managed runners now with the option to move the same setup into their own AWS, Google Cloud or Azure account",
      "overview": [
        "**WarpBuild** provides runners for GitHub Actions in two forms. Its cloud runners are hosted by WarpBuild: after installing the GitHub App on an organization, a team changes `runs-on` to a label such as `warp-ubuntu-latest-x64-4x`. Each runner is a virtual machine created on demand and destroyed after the build. Linux x64 and arm64, Windows x64 and Apple Silicon macOS runners are offered, and concurrency on Linux and Windows is documented as unlimited.",
        "The second form is [Bring Your Own Cloud](https://www.warpbuild.com/docs/ci/byoc): WarpBuild provisions and manages runners inside your AWS, Google Cloud or Azure account for a per-minute fee, with support for custom images, spot instances and static IPs. Around the runners it offers a cache action, remote Docker builders, snapshot runners that boot from a saved disk image, and observability with OTLP export. Static IPs are a BYOC-only feature, and the GitHub App cannot be installed on personal accounts."
      ],
      "notable": [
        "Cloud spot runners and Windows 2 vCPU runners were removed on 8 June 2026, according to the cloud runner docs.",
        "macOS 13 runners were removed on 8 June 2026.",
        "From 1 October 2026 the 6 vCPU macOS runners have 14 GB RAM and the 12 vCPU runners 28 GB, with prices unchanged."
      ],
      "pros": [
        "You want the choice between vendor-hosted runners and runners inside your own AWS, Google Cloud or Azure account",
        "You want each runner in its own VM that is destroyed after the build",
        "You need Linux x64 and arm64, Windows and macOS runners under one account",
        "You want to boot jobs from a saved VM disk snapshot for incremental builds",
        "You need to export runner metrics and logs to your own OTLP backend"
      ],
      "cons": [
        "Static IPs are available only with BYOC deployments",
        "WarpBuild Cache is not supported on Windows runners",
        "Snapshot runners work only on cloud Ubuntu runners, not on BYOC, Windows or macOS",
        "macOS runners do not support nested virtualization or Docker"
      ],
      "verdict": "A managed GitHub Actions runner service with a bring-your-own-cloud option, suited to teams that want hosted runners with a path to running them in their own cloud account.",
      "use_cases": {
        "faster-builds": "WarpBuild advertises 2x faster builds and offers a cache action, snapshot runners and remote Docker builders (vendor claim).",
        "lower-cost": "Linux x64 cloud runners start at $0.004/min and BYOC runners carry a $0.002/min fee on top of your own cloud bill.",
        "docker-builds": "Remote Docker Builders build and push images with layer caching, and a type=gha cache endpoint is available on runners.",
        "arm64": "Linux arm64 runners from 2 to 32 vCPU start at $0.003/min, with Apple Silicon “xfast” variants.",
        "macos-ios": "macOS runners on Apple M4 Pro and M5 Pro hardware are offered with 6 or 12 vCPU.",
        "windows": "Windows Server 2022 and 2025 x64 runners from 4 to 32 vCPU are offered.",
        "byoc": "WarpBuild provisions and manages runners inside your own AWS, Google Cloud or Azure account."
      },
      "peers": [
        "blacksmith",
        "depot",
        "namespace",
        "ubicloud",
        "runs-on",
        "cirun",
        "github-hosted-runners"
      ],
      "sources": [
        "https://www.warpbuild.com/",
        "https://www.warpbuild.com/pricing",
        "https://www.warpbuild.com/docs/ci",
        "https://www.warpbuild.com/docs/ci/cloud-runners",
        "https://www.warpbuild.com/docs/ci/byoc",
        "https://www.warpbuild.com/docs/ci/byoc/aws",
        "https://www.warpbuild.com/docs/ci/security",
        "https://www.warpbuild.com/docs/ci/quick-start",
        "https://www.warpbuild.com/docs/ci/features/caching",
        "https://www.warpbuild.com/docs/ci/features/observability",
        "https://www.warpbuild.com/docs/ci/features/snapshot-runners",
        "https://www.warpbuild.com/legal/terms"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "buildjet",
      "name": "BuildJet for GitHub Actions",
      "company": "BuildJet OÜ",
      "hq": {
        "city": "Tallinn",
        "region": null,
        "country": "EE",
        "source": "https://buildjet.com/for-github-actions/docs/about/privacy-policy"
      },
      "kind": "managed",
      "category": "managed-github-actions",
      "website": "https://buildjet.com/for-github-actions",
      "docs": "https://buildjet.com/for-github-actions/docs",
      "pricing_url": "https://buildjet.com/for-github-actions/docs/about/pricing",
      "github": null,
      "tagline": "Managed GitHub Actions runners on dedicated bare metal; shut down in 2026 and no longer running jobs.",
      "ci_systems": [
        "GitHub Actions"
      ],
      "os": [
        "Linux"
      ],
      "arch": [
        "x64",
        "arm64"
      ],
      "isolation": "Each job ran in a KVM virtual machine that was wiped after the job",
      "isolation_short": "KVM VM",
      "isolation_class": "vm",
      "ephemeral": true,
      "open_source": false,
      "license": null,
      "self_host": false,
      "byoc": null,
      "clouds": null,
      "gpu": null,
      "gpu_detail": null,
      "hardware": "BuildJet described its fleet as custom dedicated bare-metal infrastructure built for CI, with 64 GB of SSD space per run; CPU models were not named.",
      "performance": "BuildJet advertised about 2x faster builds than GitHub-hosted runners (vendor claim)",
      "sizes": "2 to 32 vCPU with 8 to 64 GB RAM on x64; 2 to 32 vCPU with 3 to 48 GB on arm64",
      "max_size_short": "32 vCPU / 64 GB",
      "cache": "buildjet/cache was a drop-in replacement for actions/cache with 20 GB of free cache storage per repository per week; drop-in setup actions such as buildjet/setup-node were also provided.",
      "docker_layer_cache": null,
      "concurrency": "Default of 64 concurrent AMD vCPUs and 32 concurrent ARM vCPUs per account; extensions of 100 vCPUs were sold at $300/month",
      "job_limits": null,
      "static_ips": null,
      "ssh_access": null,
      "observability": null,
      "setup": "Teams authorised the BuildJet GitHub OAuth app and changed runs-on to a BuildJet label. New signups have not been accepted since 6 February 2026.",
      "runs_on": "buildjet-4vcpu-ubuntu-2204",
      "pricing": "Rates still shown in the docs, although the service no longer runs jobs: $0.004/min (2 vCPU), $0.008 (4 vCPU), $0.016 (8 vCPU), $0.032 (16 vCPU) and $0.048 (32 vCPU) on x64; arm64 was priced the same except $0.064/min at 32 vCPU.",
      "price_short": null,
      "free_tier": null,
      "compliance": null,
      "regions": null,
      "best_for": "Historical reference only; BuildJet advises former users to migrate to GitHub-hosted runners",
      "overview": [
        "**BuildJet for GitHub Actions** was a managed runner service: teams authorised its GitHub OAuth app and changed `runs-on` to a label such as `buildjet-4vcpu-ubuntu-2204`. Jobs ran in KVM virtual machines on what BuildJet called custom dedicated bare-metal infrastructure, and the VM was wiped after each job. It offered Ubuntu runners on x64 and arm64 from 2 to 32 vCPU, billed per minute, plus a cache action compatible with `actions/cache`.",
        "BuildJet [announced on 6 February 2026](https://buildjet.com/for-github-actions/blog/we-are-shutting-down) that the service was shutting down. New signups stopped that day and the service stopped running jobs after 31 March 2026. The announcement says GitHub Actions has gained faster hardware, larger runners and native ARM runners since BuildJet launched, and that the company decided to focus elsewhere. It recommends moving workflows to GitHub-hosted runners and replacing `buildjet/cache` with `actions/cache`. This record is kept for reference."
      ],
      "notable": [
        "On 6 February 2026 BuildJet announced that BuildJet for GitHub Actions was shutting down and stopped accepting new signups.",
        "The announcement set 31 March 2026 as the date after which BuildJet stops running jobs.",
        "From 6 February 2026 concurrency subscriptions were made free while pay-as-you-go billing continued, according to the announcement."
      ],
      "pros": [
        "You are auditing workflows that still reference buildjet-* labels and need to know what they ran on",
        "You want the vendor’s own migration guidance, which points to GitHub-hosted runners and actions/cache",
        "You need the last published sizes and per-minute rates for a historical comparison"
      ],
      "cons": [
        "Shut down: the service stopped running jobs after 31 March 2026",
        "New signups have not been accepted since 6 February 2026",
        "Workflows that use buildjet-* labels or buildjet/cache need to be migrated"
      ],
      "verdict": "A discontinued managed runner service for GitHub Actions that stopped running jobs after 31 March 2026; listed for reference only.",
      "use_cases": {},
      "peers": [
        "github-hosted-runners",
        "blacksmith",
        "ubicloud",
        "warpbuild",
        "depot",
        "namespace"
      ],
      "sources": [
        "https://buildjet.com/for-github-actions/blog/we-are-shutting-down",
        "https://buildjet.com/for-github-actions",
        "https://buildjet.com/for-github-actions/docs",
        "https://buildjet.com/for-github-actions/docs/about/pricing",
        "https://buildjet.com/for-github-actions/docs/about/security",
        "https://buildjet.com/for-github-actions/docs/runners/hardware",
        "https://buildjet.com/for-github-actions/docs/getting-started/run-your-first-workflow",
        "https://buildjet.com/for-github-actions/docs/about/terms-of-service",
        "https://buildjet.com/for-github-actions/docs/about/privacy-policy"
      ],
      "verified_on": "2026-11-22",
      "status": "discontinued",
      "status_note": "BuildJet announced on 6 February 2026 that BuildJet for GitHub Actions was shutting down: new signups stopped that day and the service stopped running jobs after 31 March 2026. Its website now states that the service has been shut down.",
      "alternatives": true
    },
    {
      "id": "ubicloud",
      "name": "Ubicloud Runners",
      "company": "Ubicloud Inc.",
      "hq": {
        "city": "San Francisco",
        "region": "California",
        "country": "US",
        "source": "https://www.ubicloud.com/about-ubicloud"
      },
      "kind": "managed",
      "category": "managed-github-actions",
      "website": "https://www.ubicloud.com",
      "docs": "https://www.ubicloud.com/docs/github-actions-integration/quickstart",
      "pricing_url": "https://www.ubicloud.com/docs/about/pricing",
      "github": "https://github.com/ubicloud/ubicloud",
      "tagline": "Managed Linux runners for GitHub Actions from an open-source cloud platform that runs on leased bare metal.",
      "ci_systems": [
        "GitHub Actions"
      ],
      "os": [
        "Linux"
      ],
      "arch": [
        "x64",
        "arm64"
      ],
      "isolation": "Each job runs in a fresh virtual machine provisioned with Linux KVM, using Cloud Hypervisor as the virtual machine monitor; the VM and its block storage are destroyed when the job finishes",
      "isolation_short": "KVM VM",
      "isolation_class": "vm",
      "ephemeral": true,
      "open_source": true,
      "license": "AGPL-3.0",
      "self_host": true,
      "byoc": null,
      "clouds": [
        "Hetzner",
        "Leaseweb",
        "AWS",
        "Bare metal"
      ],
      "gpu": null,
      "gpu_detail": null,
      "hardware": "Standard x64 runners use AMD EPYC 9454P (two vCPUs per physical core), premium x64 runners use AMD Ryzen 9 7950X3D, and arm64 runners use Ampere Altra Q80-30 with one physical core per vCPU; all have NVMe SSDs.",
      "performance": "Ubicloud says premium runners are around 2x faster on CPU-heavy workflows and that its cache gives 4x faster downloads and 3x faster uploads (vendor claims)",
      "sizes": "2 to 30 vCPU with 8 to 120 GB RAM on x64; 2 to 30 vCPU with 6 to 90 GB on arm64",
      "max_size_short": "30 vCPU / 120 GB",
      "cache": "A transparent cache backend serves actions/cache, setup-* package caching and Docker layer caching through type=gha without workflow changes. Each repository gets 30 GB free (100 GB with premium runners enabled), entries unused for seven days are removed, and extra storage is $0.04 per GB per month on request.",
      "docker_layer_cache": true,
      "concurrency": null,
      "job_limits": null,
      "static_ips": null,
      "ssh_access": true,
      "observability": null,
      "setup": "Sign up, add a credit card, install the Ubicloud Managed Runners GitHub App, then change runs-on to a label such as ubicloud-standard-2.",
      "runs_on": "ubicloud-standard-2",
      "pricing": "Standard x64 per minute: $0.00125 (2 vCPU), $0.0025 (4), $0.005 (8), $0.01 (16) and $0.01875 (30 vCPU). Premium x64: $0.002 (2 vCPU) to $0.03 (30 vCPU). arm64 is priced the same as standard x64. Billing is per minute, and new users get premium runners by default with fallback to standard runners when premium capacity is full.",
      "price_short": "$0.00125/min",
      "free_tier": "$2.5/month credit, which Ubicloud equates to 1,250 runner minutes",
      "compliance": "SOC 2 Type 2 report available through the Trust Center; GDPR data processing agreement offered",
      "regions": "Europe by default: Hetzner (Falkenstein, Germany) as the primary fleet and Leaseweb (Frankfurt), with AWS eu-central-1 as overflow; US runners are used only on explicit request",
      "best_for": "Teams on GitHub Actions that want low per-minute Linux runners in European data centers from a vendor whose cloud platform is open source",
      "overview": [
        "**Ubicloud** is an open-source cloud platform (AGPL-3.0) that runs on bare metal leased from providers such as Hetzner and Leaseweb; its managed GitHub Actions runners are one of the services built on it. A team installs the Ubicloud Managed Runners GitHub App and changes `runs-on` to a label such as `ubicloud-standard-2`. Each job gets a fresh KVM virtual machine registered as a just-in-time runner, and the VM and its storage are destroyed afterwards.",
        "Runners are Linux only, on x64 and arm64, from 2 to 30 vCPU. Two x64 tiers exist: standard runners on AMD EPYC and premium runners on AMD Ryzen 9, which new accounts get by default. A [transparent cache](https://www.ubicloud.com/docs/github-actions-integration/ubicloud-cache) works with existing cache actions. Runners are provisioned in Europe unless US runners are requested, and larger sizes can spill over to AWS capacity at peak times. The platform itself can also be self-hosted."
      ],
      "notable": [
        "Premium runners on AMD Ryzen 9 CPUs were announced on 23 June 2025.",
        "Standard runner prices rose from $0.0008 to $0.0010 per minute for new deployments on 1 May 2026 and for existing resources on 1 June 2026.",
        "Prices for standard and premium runners in Germany and Finland increased by a further 25% on 1 September 2026, according to the changelog.",
        "In October 2026 8 vCPU runners joined 16 and 30 vCPU runners in spilling over to AWS when Ubicloud’s own hardware is full."
      ],
      "pros": [
        "You want Linux runners at $0.00125 per minute for 2 vCPU, on x64 or arm64",
        "You want the underlying cloud platform to be open source and self-hostable",
        "You need jobs to run in European data centers by default",
        "You want existing actions/cache and type=gha Docker layer caching to work without workflow changes",
        "You want to SSH into a runner to debug a workflow"
      ],
      "cons": [
        "Windows and macOS runners are not documented",
        "Runner prices were raised twice in 2026, in May and September",
        "US runners are used only on explicit request; runners are otherwise provisioned in Europe",
        "Overflow capacity on AWS does not support nested virtualization (KVM)"
      ],
      "verdict": "A low-priced managed Linux runner service for GitHub Actions built on an open-source cloud, suited to teams that want x64 and arm64 runners in Europe.",
      "use_cases": {
        "lower-cost": "Standard 2 vCPU runners are listed at $0.00125 per minute against $0.006 for GitHub-hosted runners, with a $2.5 monthly credit.",
        "faster-builds": "Premium runners use AMD Ryzen 9 7950X3D CPUs, which Ubicloud says are around 2x faster on CPU-heavy workflows (vendor claim).",
        "arm64": "Ampere Altra arm64 runners from 2 to 30 vCPU are priced the same as standard x64 runners.",
        "open-source": "The Ubicloud platform that provisions the runners is published under AGPL-3.0 and can be self-hosted on bare metal."
      },
      "peers": [
        "blacksmith",
        "depot",
        "namespace",
        "warpbuild",
        "github-hetzner-runners",
        "cirun",
        "github-hosted-runners"
      ],
      "sources": [
        "https://www.ubicloud.com/use-cases/github-actions",
        "https://www.ubicloud.com/docs/about/pricing",
        "https://www.ubicloud.com/docs/github-actions-integration/quickstart",
        "https://www.ubicloud.com/docs/github-actions-integration/runner-types",
        "https://www.ubicloud.com/docs/github-actions-integration/security",
        "https://www.ubicloud.com/docs/github-actions-integration/compliance",
        "https://www.ubicloud.com/docs/github-actions-integration/ubicloud-cache",
        "https://www.ubicloud.com/docs/github-actions-integration/debug-workflow-with-ssh",
        "https://www.ubicloud.com/docs/about/regions",
        "https://www.ubicloud.com/docs/changelog/2026-08",
        "https://www.ubicloud.com/docs/changelog/2026-09",
        "https://www.ubicloud.com/docs/changelog/2026-10",
        "https://www.ubicloud.com/blog/ubicloud-price-adjustment-2026",
        "https://www.ubicloud.com/blog/ubicloud-premium-runners-40-faster-builds-10x-larger-cache",
        "https://www.ubicloud.com/about-ubicloud",
        "https://www.ubicloud.com/docs/about/privacy-policy",
        "https://github.com/ubicloud/ubicloud"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "luxor-runners",
      "name": "Luxor Runners",
      "company": "Luxor Agent Cloud, LLC (Luxor AI, a division of Luxor Technology Corporation)",
      "hq": {
        "city": "Seattle",
        "region": "Washington",
        "country": "US",
        "source": "https://luxor.ai/company/about.md",
        "entity": "Luxor Technology"
      },
      "kind": "managed",
      "category": "managed-github-actions",
      "website": "https://luxor.ai",
      "docs": "https://luxor.ai/docs",
      "pricing_url": "https://luxor.ai/docs/pricing",
      "github": null,
      "tagline": "GitHub Actions runners on bare-metal Linux x64 and Apple Silicon macOS, one single-use VM per job; called Tenki until November 2026.",
      "ci_systems": [
        "GitHub Actions"
      ],
      "os": [
        "Linux",
        "macOS"
      ],
      "arch": [
        "x64",
        "arm64"
      ],
      "isolation": "Each job runs in a single-use VM with its own kernel on bare-metal hosts and is destroyed when the job ends; the hypervisor is not named",
      "isolation_short": "Single-use VM",
      "isolation_class": "vm",
      "ephemeral": true,
      "open_source": false,
      "license": null,
      "self_host": false,
      "byoc": null,
      "clouds": null,
      "gpu": null,
      "gpu_detail": null,
      "hardware": "Linux x64 runners on bare-metal hosts that Luxor describes as “5th Gen AMD Ryzen” and says it owns and operates, with two vCPUs mapped to one physical core; macOS runners on Apple Silicon M4 Pro.",
      "performance": "Luxor says its runners are up to 3x faster than GitHub-hosted runners on its own published benchmarks (vendor claim)",
      "sizes": "Linux x64: 2 vCPU / 4 GB to 16 vCPU / 32 GB, larger on request; macOS: 2 vCPU / 4 GB to 14 vCPU / 38 GB",
      "max_size_short": "16 vCPU / 32 GB",
      "cache": "A hosted drop-in backend for the GitHub Actions cache (entries expire after 7 days, up to 10 GiB each), pull-through mirrors for npm, APT, Go modules, Cargo, Maven, CocoaPods, Swift packages and Docker Hub, and sticky disks that keep cache directories between Linux jobs. The docs say the Docker Hub mirror does not cache build layers between jobs.",
      "docker_layer_cache": null,
      "concurrency": "5 concurrent sessions on Starter and 100 on Team, shared between runner jobs and sandbox sessions; jobs above the limit queue",
      "job_limits": "No runner-side time limit; the docs say GitHub’s 6 hours per job applies. Starter is capped at 2 vCPU / 4 GB (4 vCPU / 8 GB with a payment method on file) and Team at 16 vCPU / 64 GB",
      "static_ips": true,
      "ssh_access": null,
      "observability": null,
      "setup": "Connect a GitHub organization by installing the Luxor Runner GitHub App, then change runs-on by hand or let the migration wizard open a pull request.",
      "runs_on": "luxor-ubuntu-medium-4c-8g",
      "pricing": "Linux x64 is $0.002 per vCPU-minute: $0.004/min (2 vCPU / 4 GB), $0.008 (4 vCPU), $0.016 (8 vCPU), $0.032 (16 vCPU). macOS is $0.020 per vCPU-minute, from $0.040/min (2 vCPU) to $0.280/min (14 vCPU). Cache and sticky-disk storage is $0.20 per GiB-month; the Team plan is $250/month billed monthly or $200/month billed yearly.",
      "price_short": "$0.004/min",
      "free_tier": "$100 in free credits once a payment method is added on the $0 Starter plan; Team includes $100 of credits per month",
      "compliance": "Luxor says its parent platform has completed a SOC 2 Type 2 examination and that a SOC 2 report for Luxor Agent Cloud is in progress",
      "regions": null,
      "best_for": "Teams on GitHub Actions that want per-vCPU-minute Linux x64 and Apple Silicon macOS runners with a fixed egress range",
      "overview": [
        "**Luxor Runners** is the GitHub Actions runner product of Luxor Agent Cloud, the service that was called Tenki Cloud until November 16, 2026. A team installs the GitHub App on its organization and changes `runs-on` to a `luxor-*` label; older `tenki-*` labels keep working. Each job runs in a single-use virtual machine on bare-metal Linux x64 hosts or Apple Silicon M4 Pro Macs, and the VM is destroyed when the job ends.",
        "Pricing is per vCPU-minute, billed by the second, with queue time not billed. The same workspace also holds Luxor Sandbox and Luxor Code Reviewer, and runner jobs share one concurrency limit and one storage pool with sandboxes. Runner traffic leaves from a fixed CIDR, dedicated egress addresses cost $100 per IP per month, and dedicated Linux labels can join a Tailscale tailnet. [Pricing](https://luxor.ai/docs/pricing) lists a $0 Starter plan and a Team plan."
      ],
      "notable": [
        "November 16, 2026: Tenki Cloud was renamed Luxor Agent Cloud and the company became Luxor AI; existing tenki-* runner labels and GitHub App installs keep working.",
        "December 8, 2025: macOS runners on Apple M4 Pro machines launched in waitlist-only early access (changelog).",
        "October 2026: sticky disks were enabled by default (October 2), and the 1.5x Premium Runners surcharge was removed while runner cache billing of $0.20 per GiB-month was added (October 16).",
        "October 30, 2026: dedicated CI jobs gained support for joining a saved Tailscale network."
      ],
      "pros": [
        "You want Linux x64 runners priced at a flat $0.002 per vCPU-minute, billed by the second with queue time not billed",
        "You need Apple Silicon macOS runners (M4 Pro) from the same vendor as your Linux runners",
        "You need runner traffic to leave from a fixed CIDR, or a dedicated egress IP at $100 per month",
        "You want cache directories kept between Linux jobs on sticky disks instead of archived and restored"
      ],
      "cons": [
        "No Windows runners (stated on luxor.ai); Linux arm64 runners are not documented",
        "Sticky disks do not support macOS runners or container jobs",
        "Starter plan is capped at 2 vCPU / 4 GB runners, or 4 vCPU / 8 GB with a payment method on file",
        "A SOC 2 report for Luxor Agent Cloud itself is still in progress"
      ],
      "verdict": "A managed GitHub Actions runner service on vendor-owned bare metal, renamed from Tenki to Luxor in November 2026, that suits teams wanting simple per-vCPU pricing for Linux x64 and macOS jobs.",
      "use_cases": {
        "faster-builds": "Luxor publishes benchmarks claiming up to 3x faster runs than GitHub-hosted runners, with sticky disks and package mirrors to cut repeat work (vendor claim).",
        "lower-cost": "Linux x64 runners are listed at $0.004 per minute for 2 vCPU, billed by the second, with $100 in starting credits.",
        "macos-ios": "macOS 15 and macOS 26 runners on Apple Silicon M4 Pro are offered from 2 to 14 vCPU at $0.020 per vCPU-minute."
      },
      "peers": [
        "blacksmith",
        "depot",
        "namespace",
        "warpbuild",
        "ubicloud",
        "cirrus-runners",
        "github-hosted-runners"
      ],
      "sources": [
        "https://luxor.ai/",
        "https://luxor.ai/docs/migrate-to-luxor.md",
        "https://luxor.ai/blog/tenki-is-now-luxor-agent-cloud",
        "https://luxor.ai/products/runners.md",
        "https://luxor.ai/docs/pricing",
        "https://luxor.ai/docs/runners/sizes.md",
        "https://luxor.ai/docs/runners/limits-and-concurrency.md",
        "https://luxor.ai/docs/runners/caching.md",
        "https://luxor.ai/docs/runners/sticky-disks.md",
        "https://luxor.ai/docs/runners/networking.md",
        "https://luxor.ai/docs/runners/quickstart.md",
        "https://luxor.ai/docs/trust/security.md",
        "https://luxor.ai/docs/changelog.md",
        "https://luxor.ai/company/about.md",
        "https://luxor.ai/docs/terms-of-service.md"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "cirrus-runners",
      "name": "Cirrus Runners",
      "company": "Cirrus Labs Inc",
      "hq": null,
      "kind": "managed",
      "category": "managed-github-actions",
      "website": "https://cirrus-runners.app",
      "docs": "https://cirrus-runners.app/setup/",
      "pricing_url": "https://cirrus-runners.app/pricing/",
      "github": null,
      "tagline": "Flat-rate GitHub Actions runners for macOS and Linux, priced per concurrent runner; closed to new customers since April 2026.",
      "ci_systems": [
        "GitHub Actions"
      ],
      "os": [
        "Linux",
        "macOS"
      ],
      "arch": [
        "x64",
        "arm64"
      ],
      "isolation": "Each job runs in a single-use virtual machine that is destroyed after the job, on bare-metal hosts using Cirrus Labs’ own Tart (macOS) and Vetu (Linux) virtualization",
      "isolation_short": "Single-use VM",
      "isolation_class": "vm",
      "ephemeral": true,
      "open_source": false,
      "license": null,
      "self_host": false,
      "byoc": null,
      "clouds": null,
      "gpu": true,
      "gpu_detail": "The setup page lists a Linux GPU runner with 16 vCPU, 48 GB RAM and an Nvidia RTX 4000 that counts as two concurrent runners; the pricing page describes the GPU runner as 8 vCPU and 24 GB.",
      "hardware": "Apple M4 Pro for macOS runners; a September 2025 blog post names AMD Ryzen 7950X3D hosts for Linux, on hardware Cirrus Labs says it owns.",
      "performance": "Cirrus Labs says its runners are 2-3x faster than GitHub-hosted runners (vendor claim)",
      "sizes": "Linux x64: 2 vCPU / 6 GB to 32 vCPU / 96 GB; Linux arm64: 2 vCPU / 6 GB to 8 vCPU / 24 GB; macOS: 4 vCPU / 16 GB or 8 vCPU / 32 GB",
      "max_size_short": "32 vCPU / 96 GB",
      "cache": "Omni Cache runs as an automatic sidecar and serves as an alternative to actions/cache, a Docker Buildx layer cache endpoint and a Gradle remote build cache. Each subscribed runner includes 10 GB of cache storage shared across repositories; files are deleted after 100 days.",
      "docker_layer_cache": true,
      "concurrency": "Set by the number of runners purchased; each job consumes a fraction or multiple of one runner by size (Linux x64: 2 vCPU = 0.125, 16 vCPU = 1.00, 32 vCPU = 2.00)",
      "job_limits": null,
      "static_ips": null,
      "ssh_access": null,
      "observability": "Usage insights in the Cirrus Runners dashboard",
      "setup": "Install the Cirrus Runners GitHub App on an organization, complete checkout, then set runs-on to a Cirrus image label. Cirrus Labs says it is no longer accepting new customers.",
      "runs_on": "ghcr.io/cirruslabs/ubuntu-runner-amd64:24.04-md",
      "pricing": "$150 per month per concurrent runner with no limit on minutes. One runner covers one macOS 4-vCPU, Linux x64 16-vCPU or Linux arm64 8-vCPU job at a time. A 15% discount applies to a 12-month commitment, or 50% for qualifying nonprofit projects.",
      "price_short": "$150/month per concurrent runner",
      "free_tier": "The setup page describes a 10-day free trial with billing details required up front; Cirrus Labs says it no longer accepts new customers",
      "compliance": "SOC 2 Type 2 (stated on the product homepage)",
      "regions": null,
      "best_for": "Existing Cirrus Runners customers running macOS and Linux jobs on a fixed monthly price per concurrent runner",
      "overview": [
        "**Cirrus Runners** is a hosted GitHub Actions runner service from Cirrus Labs. A team installs a GitHub App and sets `runs-on` to a Cirrus image label for macOS on Apple M4 Pro, Linux x64, Linux arm64 or a Linux GPU runner. Each job runs in a single-use virtual machine built on the company’s own Tart and Vetu virtualization tools, and the price is a flat $150 per month for each concurrent runner with no limit on minutes.",
        "On April 7, 2026 Cirrus Labs announced an agreement to join OpenAI as part of its Agent Infrastructure team. The [announcement](https://cirruslabs.org/) says the company is no longer accepting new customers for Cirrus Runners and will keep supporting the service for existing customers through their existing contract periods. The product site was still online on November 22, 2026, with its pricing and setup pages still published."
      ],
      "notable": [
        "April 7, 2026: Cirrus Labs announced an agreement to join OpenAI; Cirrus Runners stopped accepting new customers and continues for existing customers through their existing contract periods.",
        "April 7, 2026: the same announcement said Cirrus CI would shut down on June 1, 2026 and that Tart, Vetu and Orchard would be relicensed under a more permissive licence.",
        "September 3, 2025: a blog post described the expansion from macOS-only to Linux runners and the Vetu virtualization tool built on Cloud Hypervisor.",
        "January 7, 2025: M4 Pro availability for Cirrus Runners was announced."
      ],
      "pros": [
        "You are an existing customer and want a fixed $150 per month per concurrent runner with no limit on minutes",
        "You need macOS runners on Apple M4 Pro alongside Linux x64 and arm64 runners from one vendor",
        "You want Docker layer, Gradle and Actions caching through the built-in Omni Cache sidecar",
        "You need a Linux runner with an Nvidia GPU"
      ],
      "cons": [
        "Closed to new customers since the April 7, 2026 announcement",
        "Support is committed only through existing customers’ current contract periods",
        "Linux images are not fully compatible with GitHub-hosted runner images (setup page)",
        "Cache storage is 10 GB per subscribed runner, with files deleted after 100 days"
      ],
      "verdict": "A flat-rate macOS and Linux runner service for GitHub Actions that, since its vendor agreed to join OpenAI in April 2026, serves existing customers only.",
      "use_cases": {
        "macos-ios": "macOS runners on Apple M4 Pro with 4 or 8 vCPU and images that carry recent Xcode versions are documented.",
        "arm64": "Linux arm64 runners from 2 to 8 vCPU are listed, each vCPU backed by one physical core according to the pricing page.",
        "gpu": "A Linux runner with an Nvidia RTX 4000 GPU is listed on the setup page.",
        "docker-builds": "Omni Cache exposes a local endpoint that Docker Buildx can use as a layer cache."
      },
      "peers": [
        "namespace",
        "blacksmith",
        "depot",
        "warpbuild",
        "macstadium-orka",
        "tart",
        "getmac",
        "github-hosted-runners"
      ],
      "sources": [
        "https://cirruslabs.org/",
        "https://cirrus-runners.app/",
        "https://cirrus-runners.app/pricing/",
        "https://cirrus-runners.app/setup/",
        "https://cirrus-runners.app/blog/",
        "https://cirrus-runners.app/blog/2025/09/03/from-mac-only-to-linux-powerhouse-journey-to-1-ci-performance/",
        "https://cirrus-runners.app/company/terms/",
        "https://cirrus-runners.app/company/security/"
      ],
      "verified_on": "2026-11-22",
      "status": "winding-down",
      "status_note": "On April 7, 2026 Cirrus Labs announced it had entered into an agreement to join OpenAI as part of the Agent Infrastructure team. The announcement says Cirrus Labs is no longer accepting new customers for Cirrus Runners but will continue supporting the service for existing customers through their existing contract periods.",
      "status_label": "closed to new customers",
      "alternatives": true
    },
    {
      "id": "machine",
      "name": "machine.dev",
      "company": "MACHINE DOT DEV TRADING PTY LTD (trading as machine.dev)",
      "hq": {
        "city": null,
        "region": "Queensland",
        "country": "AU",
        "source": "https://machine.dev/terms-of-service/"
      },
      "kind": "managed",
      "category": "managed-github-actions",
      "website": "https://machine.dev",
      "docs": "https://machine.dev/docs/",
      "pricing_url": "https://machine.dev/pricing/",
      "github": null,
      "tagline": "GPU, x64 and arm64 runners for GitHub Actions on dedicated AWS instances, with spot or on-demand tenancy chosen in the runs-on label.",
      "ci_systems": [
        "GitHub Actions"
      ],
      "os": [
        "Linux"
      ],
      "arch": [
        "x64",
        "arm64"
      ],
      "isolation": "Each job runs on a dedicated AWS instance that registers as an ephemeral self-hosted runner and shuts down when the job finishes",
      "isolation_short": "Dedicated AWS instance",
      "isolation_class": "vm",
      "ephemeral": true,
      "open_source": false,
      "license": null,
      "self_host": false,
      "byoc": null,
      "clouds": null,
      "gpu": true,
      "gpu_detail": "NVIDIA T4G, T4, L4, A10G, L40S and RTX 6000 (up to 8 per runner) plus AWS Trainium and Inferentia2; a 4-vCPU T4G runner is listed from $0.00358/min on spot and $0.01400/min on-demand.",
      "hardware": "AWS instances: x64 (Intel/AMD) and arm64 (Graviton) CPU runners, each with a default 100 GB gp3 EBS volume (6,000 IOPS, 250 MB/s) that can be raised to 16 TB.",
      "performance": "machine.dev says a runner goes from push to running job in under 60 seconds without pre-warming (vendor claim)",
      "sizes": "CPU runners from 2 vCPU / 4 GB to 64 vCPU / 128 GB on x64 and arm64; GPU runners from 4 vCPU / 8 GB (T4G) to 192 vCPU / 2,048 GB with 8 RTX 6000",
      "max_size_short": "64 vCPU / 128 GB",
      "cache": "An Actions cache served from the same AWS region as the runner is in closed beta, available by request: 25 GiB per repository per region with least-recently-used eviction, not billed during the beta. It works with actions/cache v4+ and BuildKit type=gha without workflow changes.",
      "docker_layer_cache": null,
      "concurrency": "Pay-as-you-go: 1 concurrent runner before a purchase and 5 after; Builder 10, Developer 25, Growth 30, Pro 40, Business 50. Jobs above the cap wait in arrival order for up to six hours",
      "job_limits": "30 minutes per job for non-paying teams; paying teams can run up to GitHub’s 5-day self-hosted runner limit. A spot job interrupted while running is not re-run or resumed automatically",
      "static_ips": null,
      "ssh_access": null,
      "observability": "Per-job CPU, memory, disk, network and GPU metrics on the dashboard, plus a cost breakdown for every job",
      "setup": "Sign up, install the Machine Provisioner GitHub App on an account or organization, enable self-hosted runners for the repositories, then set runs-on to a machine/... label.",
      "runs_on": "machine/gpu=T4/tenancy=spot",
      "pricing": "Per-minute USD “from” rates that vary by region. x64 2 vCPU / 4 GB: from $0.00076 spot or $0.00297 on-demand; arm64 2 vCPU / 4 GB: from $0.00040 spot or $0.00241 on-demand; x64 64 vCPU: from $0.01320 spot or $0.09520 on-demand. EBS storage is billed separately at $0.08 per GB-month, prorated to runtime; optional monthly plans from $25 to $300 add bonus credit and higher concurrency.",
      "price_short": "$0.00297/min on-demand",
      "free_tier": "$10 of free compute on signup; before any purchase, one job at a time and 30 minutes per job",
      "compliance": null,
      "regions": "Six AWS regions: us-east-1, us-east-2, us-west-2, ca-central-1, eu-south-2 and ap-southeast-2, selectable with a regions= option in the label; GPU availability varies by type",
      "best_for": "Teams on GitHub Actions that run GPU workloads such as model tests or fine-tuning in CI, or want spot-priced x64 and arm64 runners on AWS",
      "overview": [
        "**machine.dev** provides GPU and CPU runners for GitHub Actions. After installing its GitHub App, a job selects hardware in the label itself, for example `runs-on: machine/gpu=T4/tenancy=spot` or `machine/cpu=8/architecture=arm64`. Each job gets a dedicated AWS instance running Ubuntu 22.04 that registers as an ephemeral self-hosted runner and shuts down afterwards. GPU images ship with NVIDIA drivers, CUDA and the NVIDIA Container Toolkit installed.",
        "Six NVIDIA GPU tiers, AWS Trainium and Inferentia2 accelerators, and x64 or Graviton CPU runners from 2 to 64 vCPU are offered, each as spot or on-demand, across six AWS regions. Billing is per minute of runtime in US dollars, with EBS storage charged separately. The [pricing reference](https://machine.dev/docs/reference/pricing/) gives “from” rates because spot prices follow the AWS market, and concurrency depends on the plan."
      ],
      "notable": [
        "2024: company founded by Leonard O’Sullivan and Lyndon Swan, described as remote-first (about page).",
        "June 20, 2025: the post “Why We Built machine.dev” was published.",
        "April 27, 2026: nat-zero, a Terraform module for scale-to-zero NAT instances on AWS, was open sourced.",
        "June 12, 2026: per-job instance metrics became generally available, on by default."
      ],
      "pros": [
        "You need NVIDIA GPUs in GitHub Actions jobs, from a T4G up to 8 RTX 6000 on one runner",
        "You want to choose spot or on-demand tenancy per job in the runs-on label",
        "You need to pin jobs to specific AWS regions, such as eu-south-2 for EU data residency",
        "You want x64 or Graviton arm64 CPU runners up to 64 vCPU with a configurable EBS volume"
      ],
      "cons": [
        "A spot job interrupted while running is not re-run or resumed automatically",
        "Runners share outbound IP addresses with other jobs (FAQ)",
        "The Actions cache is in closed beta and available by request",
        "All runners use Ubuntu 22.04 LTS; Windows and macOS runners are not documented"
      ],
      "verdict": "A GPU-first managed runner service for GitHub Actions built on AWS instances, suited to machine-learning teams and to anyone wanting spot-priced CPU runners.",
      "use_cases": {
        "gpu": "Six NVIDIA GPU tiers and AWS Trainium and Inferentia2 runners are listed with CUDA and drivers preinstalled.",
        "arm64": "Graviton arm64 CPU runners from 2 to 64 vCPU are selected with an architecture=arm64 option.",
        "lower-cost": "Spot tenancy is listed from $0.00076 per minute for a 2-vCPU x64 runner, with on-demand as a fallback option."
      },
      "peers": [
        "cirun",
        "runs-on",
        "namespace",
        "ubicloud",
        "warpbuild",
        "blacksmith",
        "cirrus-runners",
        "github-hosted-runners"
      ],
      "sources": [
        "https://machine.dev/",
        "https://machine.dev/pricing/",
        "https://machine.dev/docs/",
        "https://machine.dev/docs/reference/pricing/",
        "https://machine.dev/docs/platform-specifications/cpu-runners/",
        "https://machine.dev/docs/platform-specifications/regions/",
        "https://machine.dev/docs/reference/faq/",
        "https://machine.dev/docs/guides/cache/",
        "https://machine.dev/gpu-runners/",
        "https://machine.dev/cpu-runners/",
        "https://machine.dev/about-us/",
        "https://machine.dev/terms-of-service/",
        "https://machine.dev/blog/"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "runjob",
      "name": "RunJob",
      "company": "Puzl Cloud OÜ",
      "hq": {
        "city": "Tallinn",
        "region": "Harju County",
        "country": "EE",
        "source": "https://puzl.cloud/trust/"
      },
      "kind": "managed",
      "category": "managed-github-actions",
      "website": "https://runjob.eu",
      "docs": "https://runjob.eu/docs/",
      "pricing_url": "https://runjob.eu/#pricing",
      "github": null,
      "tagline": "EU-based managed runners for GitHub Actions and GitLab CI that run each job in a fresh KVM virtual machine and bill only CPU and memory used.",
      "ci_systems": [
        "GitHub Actions",
        "GitLab CI"
      ],
      "os": [
        "Linux"
      ],
      "arch": [
        "x64"
      ],
      "isolation": "Each job runs in its own dynamically spawned KVM virtual machine (a “Spike Instance”) with its own kernel; the job filesystem is erased when the job completes",
      "isolation_short": "KVM virtual machine",
      "isolation_class": "vm",
      "ephemeral": true,
      "open_source": false,
      "license": null,
      "self_host": false,
      "byoc": null,
      "clouds": null,
      "gpu": true,
      "gpu_detail": "GPU use is listed as an Enterprise-plan feature, billed per allocated GPU-second on top of any plan; models and prices are not published.",
      "hardware": null,
      "performance": "Puzl says jobs run up to 2x faster and up to 5x cheaper than on other platforms (vendor claim)",
      "sizes": "No fixed sizes: a job’s VM scales with load up to a per-job limit of 12 vCPU / 32 GB on Free and 48 vCPU / 96 GB on Business; custom on Enterprise",
      "max_size_short": "48 vCPU / 96 GB",
      "cache": "GitHub Actions jobs use RunJob’s own cache action (puzl-cloud/github-actions-cache) with 30-day default retention, or shared persistent mount points. Job cache and persistent file storage are listed for the Business and Enterprise plans. A local Docker layer cache is listed as coming soon.",
      "docker_layer_cache": false,
      "concurrency": "10 concurrent jobs on Free; unlimited on Business and Enterprise",
      "job_limits": "Ephemeral storage per job is 100 GB for GitLab and 150 GB for GitHub; GitLab jobs get up to 2 service containers on Free and 10 on Business",
      "static_ips": null,
      "ssh_access": true,
      "observability": "A cloud console lists jobs, integrations and runners; the same resources can be managed through a Kubernetes-style API and Terraform modules",
      "setup": "GitHub Actions: install the RunJob app in a GitHub organization and change runs-on to runjob-ubuntu-latest. GitLab CI: create an integration and a runner claim with a GitLab token; existing pipeline files stay unchanged.",
      "runs_on": "runjob-ubuntu-latest",
      "pricing": "Load-based: €0.00002 per utilized vCPU-second and €0.000001 per utilized GB-second of memory, with idle CPU not billed; persistent storage is €0.00000006 per GB-second. Free plan: €0/month with 400 vCPU-minutes and 800 GB-minutes included. Business: €50/month with 2,000 vCPU-minutes and 4,000 GB-minutes included. Prices exclude VAT.",
      "price_short": "€0.00002 per vCPU-second used",
      "free_tier": "Free plan at €0/month with 400 vCPU-minutes and 800 GB-minutes per month, 1 integration and 10 concurrent jobs; the Business plan has a 14-day trial",
      "compliance": "SOC 2 Type II; Puzl also states GDPR compliance",
      "regions": "Described as EU based; specific locations are not published, and choosing a run region is listed as an Enterprise feature",
      "best_for": "Teams on GitLab CI or GitHub Actions that want EU-hosted runners billed by actual CPU and memory use instead of by runner size",
      "overview": [
        "**RunJob** (formerly RunMyJob by puzl.cloud) is a managed runner service from Puzl Cloud OÜ for GitHub Actions and GitLab CI. GitHub organizations install an app and change `runs-on` to `runjob-ubuntu-latest`; GitLab users register a RunJob runner against GitLab.com or a self-hosted instance and keep their pipeline files as they are. Each job runs in a fresh KVM virtual machine that the company calls a Spike Instance.",
        "Instead of fixed runner sizes, a job’s VM draws CPU and memory as the workload needs them, up to a per-plan limit, and billing counts only the vCPU-seconds and GB-seconds actually used. The service describes itself as EU based and SOC 2 Type II compliant. Runners can be managed from a console, a Kubernetes-style declarative API or Terraform modules, and Business and Enterprise plans add an interactive web terminal for GitLab jobs."
      ],
      "notable": [
        "The docs introduction states that RunJob was formerly named RunMyJob by puzl.cloud; on November 22, 2026 the puzl.cloud product and documentation URLs redirected to runjob.eu.",
        "As of November 22, 2026 the plan table lists a Windows execution environment and a local Docker layer cache as coming soon.",
        "As of November 22, 2026 the FAQ states that Linux x86-64 is the currently supported environment for CI jobs."
      ],
      "pros": [
        "You run GitLab CI, on GitLab.com or self-hosted, and want a third-party managed runner that needs no pipeline changes",
        "You want billing by CPU-seconds and memory-seconds actually used, with idle CPU not charged",
        "You need each job isolated in its own KVM virtual machine at an EU-based, SOC 2 Type II provider",
        "You want to manage runners declaratively through an API or Terraform modules"
      ],
      "cons": [
        "Linux x86-64 is the only currently supported execution environment (FAQ)",
        "GitHub runners work with organization accounts only; personal repositories are not supported",
        "Job cache and persistent file storage are not included in the Free plan",
        "A local Docker layer cache is listed as coming soon"
      ],
      "verdict": "An EU-based managed runner service for GitLab CI and GitHub Actions with usage-metered billing and one KVM virtual machine per job, suited to teams whose jobs spend much of their time below peak load.",
      "use_cases": {
        "gitlab": "RunJob registers managed runners for GitLab.com and self-hosted GitLab without changes to existing pipelines.",
        "lower-cost": "Billing covers only utilized vCPU-seconds and GB-seconds, and a free plan includes 400 vCPU-minutes per month.",
        "security": "Each job runs in its own KVM virtual machine with an ephemeral filesystem, and Puzl states SOC 2 Type II and GDPR compliance."
      },
      "peers": [
        "gitlab-hosted-runners",
        "gitlab-runner",
        "github-hosted-runners",
        "ubicloud",
        "namespace",
        "blacksmith",
        "cirun"
      ],
      "sources": [
        "https://runjob.eu/",
        "https://runjob.eu/docs/introduction",
        "https://runjob.eu/docs/getting-started",
        "https://runjob.eu/docs/security",
        "https://runjob.eu/docs/spike-instances",
        "https://runjob.eu/docs/github-actions/github-runner-labels",
        "https://runjob.eu/docs/github-actions/github-actions-cache-and-persistent-storage",
        "https://runjob.eu/docs/gitlab-pipelines/quick-setup-of-gitlab-runner",
        "https://runjob.eu/docs/resources-and-billing/nvidia-gpu",
        "https://puzl.cloud",
        "https://puzl.cloud/pricing/",
        "https://puzl.cloud/trust/"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "buildpulse-runners",
      "name": "BuildPulse Runners",
      "company": "BuildPulse LLC",
      "hq": null,
      "kind": "managed",
      "category": "managed-github-actions",
      "website": "https://buildpulse.io/products/runners",
      "docs": "https://buildpulse.io/runners/github-actions",
      "pricing_url": "https://buildpulse.io/pricing",
      "github": null,
      "tagline": "Managed Linux runners for GitHub Actions, CircleCI and Semaphore with included cache storage and flaky-test detection on every run.",
      "ci_systems": [
        "GitHub Actions",
        "CircleCI",
        "Semaphore"
      ],
      "os": [
        "Linux"
      ],
      "arch": [
        "x64",
        "arm64"
      ],
      "isolation": "Each job runs on a fresh runner that is discarded when the job ends; the isolation boundary is not documented",
      "isolation_short": "Fresh per-job runner",
      "isolation_class": null,
      "ephemeral": true,
      "open_source": false,
      "license": null,
      "self_host": false,
      "byoc": null,
      "clouds": null,
      "gpu": null,
      "gpu_detail": null,
      "hardware": "x64 runners on AMD EPYC 9R45 (Zen 5) according to an October 2026 benchmark post; arm64 runners on Graviton; NVMe-backed Docker builder runners from 4x to 32x.",
      "performance": "BuildPulse says its 4-vCPU runners were 1.3x to 2.1x faster than GitHub’s ubuntu-latest in its own October 2026 benchmark (vendor claim)",
      "sizes": "2 vCPU / 8 GB to 64 vCPU / 256 GB on x64; the same vCPU counts on arm64",
      "max_size_short": "64 vCPU / 256 GB",
      "cache": "250 GB of cache storage is included for Docker layer, dependency and build caches that persist across runs; storage beyond that is $0.20 per GB per month. Docker builder runners keep each customer’s layer cache between builds.",
      "docker_layer_cache": true,
      "concurrency": "Governed by a per-organization CPU quota that can be raised on request; there is no per-plan limit on parallel jobs",
      "job_limits": null,
      "static_ips": true,
      "ssh_access": null,
      "observability": "Flaky-test detection and per-test timing on every run, plus per-organization usage analytics across GitHub Actions, CircleCI and Semaphore",
      "setup": "Install the BuildPulse GitHub App on an organization and change runs-on to a bp-* label. For CircleCI, set a namespaced resource_class; for Semaphore, set the agent machine type.",
      "runs_on": "bp-ubuntu-latest-x64-4x",
      "pricing": "Linux x64 per minute (list prices dated September 2026): $0.004 (2 vCPU / 8 GB), $0.008 (4 vCPU), $0.016 (8 vCPU), $0.032 (16 vCPU), $0.064 (32 vCPU), $0.128 (64 vCPU). arm64 is $0.0025 per minute at 2 vCPU. Billed per minute, rounded up per job; cache storage beyond the included 250 GB is $0.20 per GB per month.",
      "price_short": "$0.004/min",
      "free_tier": "3,000 free minutes every month, with no credit card required",
      "compliance": null,
      "regions": "BuildPulse says its services run entirely on AWS in the United States; runner regions are not listed separately",
      "best_for": "Teams on GitHub Actions, CircleCI or Semaphore that want Linux runners bundled with cache storage and flaky-test analytics",
      "overview": [
        "**BuildPulse Runners** is a managed runner service from BuildPulse, a company also known for flaky-test detection. On GitHub Actions a team installs the BuildPulse GitHub App on its organization and changes `runs-on` to a label such as `bp-ubuntu-latest-x64-4x`. Linux runners come in six sizes from 2 to 64 vCPU on x64 and arm64, and each job runs on a fresh runner that is discarded when the job ends.",
        "The same runners can serve CircleCI through a namespaced resource class and Semaphore through an agent machine type. Billing is per minute with no minimums or commitment. Every account gets 3,000 free minutes a month and 250 GB of cache storage for Docker layers, dependencies and build outputs. Each run is parsed for flaky tests and per-test timing, and runners egress from four fixed addresses, with a dedicated egress IP available on request."
      ],
      "notable": [
        "October 23, 2026: BuildPulse published a benchmark of its runners against GitHub-hosted runners, with a public benchmark repository.",
        "September 2026: the date given for the current runner list prices on the product and pricing pages.",
        "As of November 22, 2026 the security page says BuildPulse is pursuing a SOC 2 Type I attestation."
      ],
      "pros": [
        "You want one runner vendor across GitHub Actions, CircleCI and Semaphore",
        "You want 3,000 free minutes each month and 250 GB of cache storage included",
        "You want flaky-test detection and per-test timing on every run without extra setup",
        "You need runners that egress from fixed IP addresses for allowlisting"
      ],
      "cons": [
        "Not yet SOC 2 attested; the security page says a Type I attestation is being pursued",
        "Runners register with GitHub organizations, not personal accounts",
        "The isolation boundary between jobs is not documented",
        "Windows and macOS runners are not documented"
      ],
      "verdict": "A managed Linux runner service for GitHub Actions, CircleCI and Semaphore that suits teams wanting runners, caching and test analytics from one vendor.",
      "use_cases": {
        "faster-builds": "BuildPulse reports 1.3x to 2.1x faster runs than GitHub’s ubuntu-latest at 4 vCPU in its own benchmark (vendor claim).",
        "lower-cost": "Linux x64 runners are listed at $0.004 per minute for 2 vCPU, with 3,000 free minutes a month.",
        "docker-builds": "Docker layer caches persist across runs, and NVMe-backed Docker builder runners are offered.",
        "arm64": "arm64 runners on Graviton are offered in the same vCPU sizes as x64, from $0.0025 per minute at 2 vCPU."
      },
      "peers": [
        "blacksmith",
        "depot",
        "warpbuild",
        "namespace",
        "ubicloud",
        "buildjet",
        "github-hosted-runners"
      ],
      "sources": [
        "https://buildpulse.io/products/runners",
        "https://buildpulse.io/runners/github-actions",
        "https://buildpulse.io/pricing",
        "https://buildpulse.io/security",
        "https://buildpulse.io/terms",
        "https://buildpulse.io/blog/github-actions-vs-buildpulse-runners-benchmark",
        "https://platform.buildpulse.io/docs"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "avrea",
      "name": "Avrea",
      "company": "Avrea Oy",
      "hq": {
        "city": "Helsinki",
        "region": null,
        "country": "FI",
        "source": "https://avrea.com/terms"
      },
      "kind": "managed",
      "category": "managed-github-actions",
      "website": "https://avrea.com",
      "docs": "https://docs.avrea.com/",
      "pricing_url": "https://avrea.com/pricing",
      "github": null,
      "tagline": "Managed GitHub Actions runners for Linux, macOS and Windows on dedicated hardware, with colocated caches and SSH into running jobs.",
      "ci_systems": [
        "GitHub Actions"
      ],
      "os": [
        "Linux",
        "macOS",
        "Windows"
      ],
      "arch": [
        "x64",
        "arm64"
      ],
      "isolation": "Every CI job runs in its own VM on dedicated hardware, destroyed when the job ends; the hypervisor is not named",
      "isolation_short": "Per-job VM",
      "isolation_class": "vm",
      "ephemeral": true,
      "open_source": false,
      "license": null,
      "self_host": false,
      "byoc": null,
      "clouds": null,
      "gpu": null,
      "gpu_detail": null,
      "hardware": "Dedicated bare-metal AMD EPYC 4585PX hosts with DDR5 memory and local NVMe storage for x86 runners; Apple M5 Max machines for macOS and ARM Linux runners.",
      "performance": "Avrea says CI runs 2x faster on hardware alone and up to 10x faster with cache hits (vendor claim)",
      "sizes": "Linux x86: 1 vCPU / 4 GB to 32 vCPU / 128 GB; Linux ARM: 1 vCPU / 4 GB to 16 vCPU / 64 GB; macOS: 8 vCPU / 24 GB or 16 vCPU / 48 GB; Windows: 2 vCPU / 8 GB to 16 vCPU / 64 GB",
      "max_size_short": "32 vCPU / 128 GB",
      "cache": "Colocated caches cover the GitHub Actions cache, build outputs, package downloads, Git checkouts and LFS data, with 25 GB per repository included. Docker layers are cached by pointing BuildKit’s type=gha backend at Avrea’s cache, and optional persistent disks can keep Docker state between Linux jobs.",
      "docker_layer_cache": true,
      "concurrency": "Caps on concurrent jobs and vCPUs depend on usage and rise as usage grows; no figures are published",
      "job_limits": null,
      "static_ips": true,
      "ssh_access": true,
      "observability": "Flake analytics, run history, step-level logs and runner size suggestions based on each job’s CPU, memory and disk use",
      "setup": "Install the Avrea GitHub App on a GitHub organization or personal account, then change runs-on from ubuntu-latest to avrea-ubuntu-latest.",
      "runs_on": "avrea-ubuntu-latest-4-vcpu",
      "pricing": "Per-minute rates, billed by the second. Linux x86: $0.002 (1 vCPU / 4 GB), $0.004 (2 vCPU / 8 GB), $0.008 (4 vCPU / 16 GB), $0.016 (8 vCPU / 32 GB), $0.032 (16 vCPU), $0.064 (32 vCPU). Linux ARM from $0.006 (1 vCPU) to $0.096 (16 vCPU); macOS $0.08 (8 vCPU) and $0.16 (16 vCPU); Windows $0.008 (2 vCPU) to $0.064 (16 vCPU). Extra cache is about $0.50 per GB per month.",
      "price_short": "$0.004/min",
      "free_tier": "$12 of credit each month on the pay-as-you-go plan, which Avrea says covers 3,000 minutes on a 2-vCPU Linux x86 runner",
      "compliance": "ISO/IEC 27001:2022 certified; SOC 2 Type II attested",
      "regions": null,
      "best_for": "Teams on GitHub Actions that need Linux, macOS and Windows runners from one vendor, with built-in caching and SSH debugging",
      "overview": [
        "**Avrea** is a Helsinki-registered company that offers managed runners for GitHub Actions. A team installs the Avrea GitHub App on an organization or personal account and changes `runs-on` to a label such as `avrea-ubuntu-latest`. Every job runs in its own virtual machine on dedicated hardware and the VM is destroyed when the job ends. Runners are offered for Linux on x86 and ARM, for macOS on Apple Silicon and for Windows 2025.",
        "Avrea places caches next to the runners for the Actions cache, package downloads, Docker layers and Git data, and scopes them to each repository. SSH access to a running job is on by default from the console or the `avr` CLI, and an egress firewall limits which hosts runner VMs can reach. Billing is pay as you go with a $12 monthly credit; dedicated static IP egress is listed as an Enterprise feature."
      ],
      "notable": [
        "26 May 2026: Avrea announced its launch and $4.7 million in pre-seed funding led by Earlybird.",
        "13 May 2026 and 3 June 2026: ISO 27001:2022 certification and SOC 2 Type II attestation were announced.",
        "9 June 2026: macOS and ARM Linux runners on Apple M5 Max were announced.",
        "27 October 2026: persistent disks for runners were announced."
      ],
      "pros": [
        "You need Linux x86, Linux ARM, macOS and Windows runners under one account and one label scheme",
        "You want to SSH into a running job from a browser or CLI without changing the workflow",
        "You want Actions, package, Docker layer and Git caches served from storage colocated with the runners",
        "You need an egress firewall per organization and repository, with a record of each job’s connections",
        "You need a vendor with ISO/IEC 27001:2022 certification and a SOC 2 Type II attestation"
      ],
      "cons": [
        "Runner and data center locations are not published; the trust page says to contact Avrea about data location requirements",
        "Dedicated static IP egress is listed as an Enterprise feature at $100 per IP per month",
        "Linux ARM runners are listed at $0.012 per minute for 2 vCPU, against $0.004 for the same size on x86",
        "Concurrency caps are not published and start lower for free-tier organizations"
      ],
      "verdict": "A managed GitHub Actions runner service on dedicated hardware that covers Linux, macOS and Windows and suits teams that want caching, SSH debugging and egress controls included.",
      "use_cases": {
        "faster-builds": "Avrea claims 2x faster runs from its hardware and larger gains from colocated caches (vendor claim).",
        "docker-builds": "Docker layers are cached through BuildKit’s type=gha backend served by Avrea, with optional persistent-disk Docker state on Linux.",
        "macos-ios": "macOS 26 and 27 runners on Apple M5 Max are offered with 8 or 16 vCPU.",
        "windows": "Windows 2025 runners are offered from 2 to 16 vCPU.",
        "security": "Per-job VMs, repository-scoped caches, an egress firewall, ISO/IEC 27001:2022 certification and a SOC 2 Type II attestation are documented."
      },
      "peers": [
        "blacksmith",
        "depot",
        "namespace",
        "warpbuild",
        "ubicloud",
        "luxor-runners",
        "github-hosted-runners"
      ],
      "sources": [
        "https://avrea.com",
        "https://avrea.com/pricing",
        "https://docs.avrea.com/",
        "https://docs.avrea.com/runners/",
        "https://docs.avrea.com/getting-started/",
        "https://docs.avrea.com/security/",
        "https://docs.avrea.com/debugging/",
        "https://docs.avrea.com/static-ip/",
        "https://docs.avrea.com/cache/build-cache/docker/",
        "https://avrea.com/trust-security",
        "https://avrea.com/about-us",
        "https://avrea.com/terms",
        "https://avrea.com/blog",
        "https://avrea.com/blog/launching-avrea"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "craci",
      "name": "CRACI",
      "company": "CRACI Corporation Oy",
      "hq": {
        "city": "Helsinki",
        "region": null,
        "country": "FI",
        "source": "https://craci.com/privacy-policy"
      },
      "kind": "managed",
      "category": "managed-github-actions",
      "website": "https://craci.com",
      "docs": "https://docs.craci.com",
      "pricing_url": "https://craci.com/pricing",
      "github": null,
      "tagline": "Managed Linux runners for GitHub Actions on European bare metal that record a build-time SBOM and apply an egress policy to each job.",
      "ci_systems": [
        "GitHub Actions"
      ],
      "os": [
        "Linux"
      ],
      "arch": [
        "x64",
        "arm64"
      ],
      "isolation": "Each job runs in its own isolated virtual machine on bare-metal infrastructure; the hypervisor is not named",
      "isolation_short": "Per-job VM",
      "isolation_class": "vm",
      "ephemeral": null,
      "open_source": false,
      "license": null,
      "self_host": false,
      "byoc": null,
      "clouds": null,
      "gpu": null,
      "gpu_detail": null,
      "hardware": "European bare-metal infrastructure in data centers in Finland and Sweden; CPU models are not published.",
      "performance": "CRACI says jobs run twice as fast as on GitHub-hosted runners (vendor claim)",
      "sizes": "Any size from 1 to 32 vCPU with 3 GiB of memory per vCPU on Starter (default 2 vCPU / 6 GiB); up to 64 vCPU on Secure and Enterprise",
      "max_size_short": "32 vCPU / 96 GB",
      "cache": "Autocaching for common public packages is included, with private sources on the Secure plan. Other cache features are not documented; the docs note that an SBOM may be incomplete when a job restores dependencies from a cache.",
      "docker_layer_cache": null,
      "concurrency": "No limit on parallel jobs on any plan; while running on free credits, jobs can use 8 vCPUs in total at a time",
      "job_limits": null,
      "static_ips": null,
      "ssh_access": null,
      "observability": "Per-job SBOMs, network traces and provenance records, retained for 180 days by default",
      "setup": "Connect a GitHub organization by installing the CRACI GitHub App, then change runs-on to craci. Size, image and network policy are set with the cracicorp/setup@v1 action or with runner templates.",
      "runs_on": "craci",
      "pricing": "€0.002 per vCPU-minute, billed by the second, at the same rate on x86-64 and ARM64: €0.004/min (2 vCPU / 6 GB), €0.008 (4 vCPU), €0.016 (8 vCPU), €0.032 (16 vCPU), €0.064 (32 vCPU). Starter has no platform fee; Secure is €2,000/month plus usage with 10% off compute.",
      "price_short": "€0.004/min",
      "free_tier": "€10 of free credit on the Starter plan, with no card needed",
      "compliance": "ISO/IEC 27001 (certificate issued by Into Certification)",
      "regions": "Data centers in Finland and Sweden",
      "best_for": "Teams on GitHub Actions that want European-hosted Linux runners with per-job SBOMs and network egress policy",
      "overview": [
        "**CRACI** is a Helsinki-based company whose managed runners replace the GitHub-hosted runner while workflows, reviews and deployments stay in GitHub Actions. A team installs the CRACI GitHub App on its organization and sets `runs-on: craci`. Jobs run in isolated virtual machines on bare-metal servers in Finland and Sweden, on Ubuntu 24.04 images for x64 or ARM64. Only Linux jobs are supported.",
        "The product is built around supply-chain evidence: CRACI records the packages a job downloads into an SBOM without a plugin, keeps network traces and provenance records, and checks an egress policy before a job starts. Runner size is a number of compute units, each 1 vCPU and 3 GiB of memory, and compute costs €0.002 per vCPU-minute on either architecture, billed by the second. The Starter plan has no platform fee."
      ],
      "notable": [
        "2025: CRACI was founded in Helsinki (about page).",
        "2026: a €1.4M pre-seed round led by Lifeline Ventures, with First Fellow Partners and Wave Ventures, was announced (about page).",
        "As of November 22, 2026 the pricing page says early-adopter terms stay in place through 2027."
      ],
      "pros": [
        "You need CI jobs to run on bare-metal infrastructure in Finland and Sweden",
        "You want an SBOM and network trace recorded for every job without adding workflow steps",
        "You want to restrict outbound network access from jobs with a policy checked before the job starts",
        "You want x86-64 and ARM64 runners at the same per-vCPU rate, with no limit on parallel jobs"
      ],
      "cons": [
        "Runs Linux jobs only; macOS and Windows jobs stay on other runners (quickstart)",
        "GitHub Actions is the only CI system supported today",
        "While on free credits, jobs can use 8 vCPUs in total at a time",
        "An SBOM may be incomplete when a job restores dependencies from a cache"
      ],
      "verdict": "A European managed runner service for GitHub Actions that pairs per-job virtual machines with SBOM, provenance and egress-policy features, suited to teams with supply-chain or data-location requirements.",
      "use_cases": {
        "security": "Each job gets an isolated VM, a recorded SBOM and network trace, and an egress policy checked before it starts; CRACI states ISO/IEC 27001 certification.",
        "arm64": "A full Ubuntu 24.04 image is available for native ARM64 builds at the same per-vCPU rate as x86-64.",
        "lower-cost": "Compute is €0.002 per vCPU-minute billed by the second, with no platform fee on Starter and €10 of free credit."
      },
      "peers": [
        "blacksmith",
        "depot",
        "namespace",
        "ubicloud",
        "warpbuild",
        "runjob",
        "actuated",
        "github-hosted-runners"
      ],
      "sources": [
        "https://craci.com",
        "https://craci.com/pricing",
        "https://docs.craci.com",
        "https://docs.craci.com/quickstart/",
        "https://docs.craci.com/reference/runner-configuration/",
        "https://craci.com/code-safety",
        "https://craci.com/security",
        "https://craci.com/about",
        "https://craci.com/terms-of-service",
        "https://craci.com/privacy-policy"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "actions-runner-controller",
      "name": "Actions Runner Controller (ARC)",
      "company": "GitHub, Inc.",
      "hq": {
        "city": "San Francisco",
        "region": "California",
        "country": "US",
        "source": "https://docs.github.com/en/site-policy/privacy-policies/github-general-privacy-statement"
      },
      "kind": "open-source",
      "category": "open-source-autoscalers",
      "website": "https://github.com/actions/actions-runner-controller",
      "docs": "https://docs.github.com/en/actions/concepts/runners/actions-runner-controller",
      "pricing_url": null,
      "github": "https://github.com/actions/actions-runner-controller",
      "tagline": "Kubernetes operator that scales self-hosted GitHub Actions runners as ephemeral pods through runner scale sets.",
      "ci_systems": [
        "GitHub Actions"
      ],
      "os": [
        "Linux",
        "Windows"
      ],
      "arch": null,
      "isolation": "Each runner is a pod in your Kubernetes cluster; jobs that need containers use Docker-in-Docker, which requires privileged mode, or Kubernetes mode, which starts separate job pods",
      "isolation_short": "Kubernetes pod",
      "isolation_class": "container",
      "ephemeral": true,
      "open_source": true,
      "license": "Apache-2.0",
      "self_host": true,
      "byoc": true,
      "clouds": [
        "Kubernetes"
      ],
      "gpu": null,
      "gpu_detail": null,
      "hardware": null,
      "performance": null,
      "sizes": null,
      "max_size_short": null,
      "cache": null,
      "docker_layer_cache": null,
      "concurrency": "Set per runner scale set with minRunners and maxRunners; with both unset, runners follow the number of assigned jobs and scale down to zero when idle",
      "job_limits": "The GitHub Actions service unassigns a job if no runner accepts it within 24 hours",
      "static_ips": null,
      "ssh_access": null,
      "observability": "The controller-manager and listener pods emit metrics in Prometheus format (from gha-runner-scale-set 0.5.0)",
      "setup": "Install the controller and one or more runner scale sets with Helm, authenticating with a GitHub App or a personal access token. The scale set installation name is the value workflows use in runs-on.",
      "runs_on": "arc-runner-set",
      "pricing": "Free and open source; you pay for the infrastructure it runs on.",
      "price_short": null,
      "free_tier": null,
      "compliance": null,
      "regions": null,
      "best_for": "Teams that already operate Kubernetes and want the GitHub-supported way to autoscale self-hosted GitHub Actions runners",
      "overview": [
        "**Actions Runner Controller (ARC)** is a Kubernetes operator that orchestrates and scales self-hosted runners for GitHub Actions. In its runner scale set mode, a listener pod waits for job messages from the GitHub Actions service, and the controller creates ephemeral runner pods that register with a just-in-time configuration token and are removed after their job. Scale sets are installed with Helm and can serve a repository, an organization or an enterprise.",
        "The project has two modes. GitHub [supports only](https://docs.github.com/en/actions/hosting-your-own-runners/managing-self-hosted-runners-with-actions-runner-controller/about-support-for-actions-runner-controller) the autoscaling runner scale sets (the `actions.github.com` API group). The older pull-driven and webhook-driven modes (`actions.summerwind.net`) are called legacy in the README and are [maintained by the community only](https://github.com/actions/actions-runner-controller/discussions/2775). Windows runners are documented only for the legacy mode. GitHub documents ARC on Kubernetes and OpenShift, and on GitHub Enterprise Server 3.9 and later."
      ],
      "notable": [
        "A repository discussion dated 25 July 2023 sets out that GitHub maintains and supports the autoscaling runner scale sets mode, while the earlier scaling modes continue under community maintenance",
        "Release gha-runner-scale-set-0.15.1 was published on 23 October 2026",
        "The 0.15 release notes also list experimental Helm charts, which they say are not supported for production workloads"
      ],
      "pros": [
        "You already run Kubernetes and want runners managed as pods and installed with Helm",
        "You want the autoscaling mode that GitHub itself maintains and supports",
        "You want runners that scale to zero when no jobs are assigned",
        "You need runners at repository, organization or enterprise level, including on GitHub Enterprise Server 3.9 or later"
      ],
      "cons": [
        "Docker-in-Docker mode requires privileged containers",
        "Helm does not manage CRD upgrades, so upgrading means uninstalling the scale sets and the controller and reinstalling them",
        "GitHub Support treats Kubernetes setup, networking, storage provisioning and image building as out of scope",
        "The legacy scaling modes receive community support only"
      ],
      "verdict": "A Kubernetes operator, maintained together with GitHub, that autoscales ephemeral runner pods; it suits teams that already operate a cluster.",
      "use_cases": {
        "open-source": "The controller is Apache-2.0 licensed and you install it in your own cluster with Helm.",
        "byoc": "Runners are pods in a Kubernetes or OpenShift cluster that you operate."
      },
      "peers": [
        "terraform-aws-github-runner",
        "garm",
        "fireactions",
        "runs-on",
        "actuated",
        "aws-codebuild-runners",
        "github-hosted-runners"
      ],
      "sources": [
        "https://github.com/actions/actions-runner-controller",
        "https://github.com/actions/actions-runner-controller/releases",
        "https://github.com/actions/actions-runner-controller/discussions/2775",
        "https://docs.github.com/en/actions/concepts/runners/actions-runner-controller",
        "https://docs.github.com/en/actions/tutorials/use-actions-runner-controller/deploy-runner-scale-sets",
        "https://docs.github.com/en/actions/tutorials/use-actions-runner-controller/quickstart",
        "https://docs.github.com/en/actions/hosting-your-own-runners/managing-self-hosted-runners-with-actions-runner-controller/about-support-for-actions-runner-controller",
        "https://github.com/actions/actions-runner-controller/blob/master/docs/configuring-windows-runners.md",
        "https://github.com/actions/actions-runner-controller/blob/master/LICENSE",
        "https://docs.github.com/en/site-policy/privacy-policies/github-general-privacy-statement"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "terraform-aws-github-runner",
      "name": "terraform-aws-github-runner",
      "company": "GitHub AWS Runners (community project)",
      "hq": null,
      "kind": "open-source",
      "category": "open-source-autoscalers",
      "website": "https://github-aws-runners.github.io/terraform-aws-github-runner/",
      "docs": "https://github-aws-runners.github.io/terraform-aws-github-runner/",
      "pricing_url": null,
      "github": "https://github.com/github-aws-runners/terraform-aws-github-runner",
      "tagline": "Terraform module that runs GitHub Actions runners on EC2 spot or on-demand instances, scaled by Lambda functions.",
      "ci_systems": [
        "GitHub Actions"
      ],
      "os": [
        "Linux",
        "Windows",
        "macOS"
      ],
      "arch": [
        "x64",
        "arm64"
      ],
      "isolation": "Each runner is an EC2 instance in your AWS account; by default runners are reused until they are idle, and in ephemeral mode each instance takes one job with a just-in-time configuration and is then terminated",
      "isolation_short": "EC2 instance",
      "isolation_class": "vm",
      "ephemeral": null,
      "open_source": true,
      "license": "MIT",
      "self_host": true,
      "byoc": true,
      "clouds": [
        "AWS"
      ],
      "gpu": true,
      "gpu_detail": "With dynamic labels enabled, ghr-ec2 override labels can request EC2 instance requirements that include GPU accelerator types.",
      "hardware": "EC2 instance types that you list, spot or on-demand, launched through the EC2 CreateFleet API; Graviton types for arm64 and EC2 Mac dedicated hosts for the experimental macOS support",
      "performance": null,
      "sizes": null,
      "max_size_short": null,
      "cache": null,
      "docker_layer_cache": null,
      "concurrency": "The scale-up Lambda stops creating instances once the configured maximum number of runners is reached; an optional pool keeps a scheduled minimum of ephemeral org-level runners",
      "job_limits": null,
      "static_ips": null,
      "ssh_access": true,
      "observability": "Lambda and runner logs go to CloudWatch; Lambda tracing can be enabled and CloudWatch metrics are available as an experimental feature",
      "setup": "Create and configure a GitHub App, download or build the Lambda artifacts, and deploy the module with Terraform. Then install the App on your organization or repositories.",
      "runs_on": "[self-hosted, linux]",
      "pricing": "Free and open source; you pay for the infrastructure it runs on.",
      "price_short": null,
      "free_tier": null,
      "compliance": null,
      "regions": null,
      "best_for": "Teams on AWS that manage infrastructure with Terraform and want GitHub Actions runners on EC2 that scale to zero",
      "overview": [
        "**terraform-aws-github-runner** is a Terraform module that creates the AWS infrastructure for self-hosted, autoscaling GitHub Actions runners on EC2. An API Gateway endpoint receives `workflow_job` webhook events, a queue buffers them, and Lambda functions launch instances through the EC2 CreateFleet API and remove them when they are idle. Runners scale down to zero when no workflows are active. Spot instances are the documented default lifecycle, with on-demand as an option.",
        "Runners are reused until idle by default; the docs recommend the optional ephemeral mode, in which an instance takes a single job using a just-in-time configuration. The module covers Linux on x64 and arm64, Windows, and experimental macOS on EC2 Mac dedicated hosts, at organization or repository level. It works with github.com, GitHub Enterprise Cloud with data residency and GitHub Enterprise Server. The project was founded at Philips Labs and now lives in the github-aws-runners organisation."
      ],
      "notable": [
        "Version 8.0.0 (21 October 2026) introduced an experimental multi-runner v2 architecture and experimental scale-set orchestration that runs a controller on ECS Fargate",
        "Version 8.1.1 was released on 12 November 2026",
        "The README states the repository was originally founded and maintained by Philips Labs; the github-aws-runners organisation describes itself as formerly hosted via philips-labs"
      ],
      "pros": [
        "You want runners on EC2 spot instances that scale down to zero when no workflows are active",
        "You manage AWS with Terraform and want the whole runner stack defined as a module",
        "You need Linux x64, Linux arm64 and Windows runners from one deployment through the multi-runner module",
        "You want to bring your own AMI, instance types and subnets",
        "You use GitHub Enterprise Server or GitHub Enterprise Cloud with data residency"
      ],
      "cons": [
        "Enterprise-level runners are not supported; runners are registered at organization or repository level",
        "The maintainers describe the project as maintained on a best-effort basis",
        "The module states it is not certified by any security organisation and that example AMIs are not hardened",
        "macOS runners, metrics and scale-set orchestration are marked experimental"
      ],
      "verdict": "A community-maintained Terraform module for webhook-driven GitHub Actions runners on EC2; it suits AWS teams that are comfortable operating Lambda, SQS and Terraform themselves.",
      "use_cases": {
        "open-source": "The module and its Lambda functions are MIT licensed and deployed entirely in your own AWS account.",
        "byoc": "Runners are EC2 instances in your own VPC and subnets.",
        "lower-cost": "Runners use EC2 spot instances by default and scale to zero when no workflows are active.",
        "arm64": "Graviton instance types are supported by setting the runner architecture to arm64.",
        "windows": "Windows is a supported runner OS, with Windows Server 2022 as the default AMI."
      },
      "peers": [
        "actions-runner-controller",
        "runs-on",
        "garm",
        "aws-codebuild-runners",
        "cirun",
        "hyperenv",
        "github-hetzner-runners"
      ],
      "sources": [
        "https://github.com/github-aws-runners/terraform-aws-github-runner",
        "https://github.com/github-aws-runners/terraform-aws-github-runner/releases",
        "https://github.com/github-aws-runners/terraform-aws-github-runner/blob/main/CHANGELOG.md",
        "https://github.com/github-aws-runners/terraform-aws-github-runner/blob/main/LICENSE.md",
        "https://github-aws-runners.github.io/terraform-aws-github-runner/",
        "https://github-aws-runners.github.io/terraform-aws-github-runner/getting-started/",
        "https://github-aws-runners.github.io/terraform-aws-github-runner/configuration/",
        "https://github-aws-runners.github.io/terraform-aws-github-runner/security/"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "garm",
      "name": "GARM (GitHub Actions Runner Manager)",
      "company": "Cloudbase Solutions",
      "hq": {
        "city": "Timișoara",
        "region": null,
        "country": "RO",
        "source": "https://cloudbase.it/about/"
      },
      "kind": "open-source",
      "category": "open-source-autoscalers",
      "website": "https://github.com/cloudbase/garm",
      "docs": "https://github.com/cloudbase/garm/blob/main/doc/README.md",
      "pricing_url": null,
      "github": "https://github.com/cloudbase/garm",
      "tagline": "Self-hosted runner manager for GitHub and Gitea Actions that creates ephemeral runners across clouds through provider plugins.",
      "ci_systems": [
        "GitHub Actions",
        "Forgejo/Gitea Actions"
      ],
      "os": [
        "Linux",
        "Windows"
      ],
      "arch": [
        "x64",
        "arm64"
      ],
      "isolation": "Runners are instances created by an external provider (cloud VMs, LXD or Incus instances, or Kubernetes pods), so the boundary depends on the provider you choose",
      "isolation_short": "Depends on provider",
      "isolation_class": "mixed",
      "ephemeral": true,
      "open_source": true,
      "license": "Apache-2.0",
      "self_host": true,
      "byoc": true,
      "clouds": [
        "AWS",
        "Azure",
        "Google Cloud",
        "Oracle Cloud",
        "OpenStack",
        "Kubernetes",
        "LXD",
        "Incus",
        "CloudStack",
        "Akamai/Linode"
      ],
      "gpu": null,
      "gpu_detail": null,
      "hardware": null,
      "performance": null,
      "sizes": null,
      "max_size_short": null,
      "cache": null,
      "docker_layer_cache": null,
      "concurrency": "Each pool or scale set has a maximum number of runners and a minimum number of idle runners",
      "job_limits": null,
      "static_ips": null,
      "ssh_access": true,
      "observability": "Prometheus metrics with ready-made Grafana dashboards and starter alert rules, plus a built-in web UI",
      "setup": "Run the GARM binary with Docker or systemd, add credentials for your forge and install a provider. Then create a pool or scale set for a repository, organization or enterprise.",
      "runs_on": "[ubuntu, generic]",
      "pricing": "Free and open source; you pay for the infrastructure it runs on.",
      "price_short": null,
      "free_tier": null,
      "compliance": null,
      "regions": null,
      "best_for": "Teams that want one self-hosted controller for GitHub or Gitea runners spread over several clouds or private infrastructure",
      "overview": [
        "**GARM** (GitHub Actions Runner Manager) is a self-hosted runner manager for GitHub Actions and Gitea Actions. A single controller creates, scales and destroys ephemeral runner instances through external provider executables. It offers two scaling modes: pools, which react to `workflow_job` webhooks and match jobs to pools by label, and scale sets, which use GitHub’s runner scale set message queue so that no webhook endpoint is needed.",
        "GARM ships as one binary with an embedded SQLite database and a built-in web UI. The README lists providers for Amazon EC2, Azure, GCP, Oracle OCI, OpenStack, LXD, Incus, Kubernetes, CloudStack and (experimental) Akamai/Linode, and providers can be written in any language. It works with github.com, GitHub Enterprise Server and Gitea, at repository, organization and enterprise level. A Kubernetes operator is maintained separately by Mercedes-Benz."
      ],
      "notable": [
        "v0.2.0 (26 April 2026) was the first stable release of the v0.2 branch, adding runner scale sets, Gitea support, agent mode with remote shell access and a new web UI",
        "v0.2.1 was released on 8 May 2026 to follow the renaming of Gitea’s runner assets",
        "The README credits Mercedes-Benz for the GARM Kubernetes operator and links its Kubernetes provider"
      ],
      "pros": [
        "You want one controller that manages runners in several clouds and on private infrastructure at once",
        "You want a single binary with an embedded SQLite database instead of an external database or message broker",
        "You run Gitea or GitHub Enterprise Server as well as github.com",
        "You want to choose between webhook-driven pools and GitHub runner scale sets",
        "You want to write your own provider for infrastructure that is not covered"
      ],
      "cons": [
        "Pools rely on workflow_job webhooks, so the forge must be able to reach GARM; the release notes cite lost webhook deliveries as a risk that scale sets avoid",
        "Scale set labels can only be set at creation time and cannot be updated afterwards",
        "The README warns that the main branch and its documentation describe unreleased code and are not guaranteed to be stable",
        "The Kubernetes, CloudStack and Linode providers are maintained outside the cloudbase organisation, and the Linode provider is marked experimental"
      ],
      "verdict": "A single-binary, multi-cloud runner manager for GitHub and Gitea Actions; it suits teams that run their own infrastructure across more than one provider.",
      "use_cases": {
        "open-source": "GARM is Apache-2.0 licensed and runs as a single self-hosted binary.",
        "byoc": "Providers create runners in your own cloud accounts, OpenStack, LXD, Incus or Kubernetes.",
        "windows": "Pools accept windows as an OS type alongside linux.",
        "arm64": "Pools accept arm64 as a runner architecture."
      },
      "peers": [
        "actions-runner-controller",
        "terraform-aws-github-runner",
        "myshoes",
        "fireactions",
        "cirun",
        "runs-on",
        "github-hetzner-runners"
      ],
      "sources": [
        "https://github.com/cloudbase/garm",
        "https://github.com/cloudbase/garm/releases",
        "https://github.com/cloudbase/garm/releases/tag/v0.2.0",
        "https://github.com/cloudbase/garm/blob/main/doc/pools-and-scaling.md",
        "https://github.com/cloudbase/garm/blob/main/doc/scale-sets.md",
        "https://github.com/cloudbase/garm/blob/main/doc/agent-and-object-store.md",
        "https://github.com/cloudbase/garm/blob/main/doc/monitoring.md",
        "https://github.com/cloudbase/garm/blob/v0.2.1/doc/managing-entities.md",
        "https://github.com/cloudbase/garm/blob/main/LICENSE",
        "https://cloudbase.it/about/"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "gitlab-runner",
      "name": "GitLab Runner",
      "company": "GitLab Inc.",
      "hq": null,
      "kind": "open-source",
      "category": "open-source-autoscalers",
      "website": "https://docs.gitlab.com/runner/",
      "docs": "https://docs.gitlab.com/runner/",
      "pricing_url": null,
      "github": "https://gitlab.com/gitlab-org/gitlab-runner",
      "tagline": "The open-source agent that runs GitLab CI/CD jobs, with executors for Docker, Kubernetes, shell and autoscaled cloud instances.",
      "ci_systems": [
        "GitLab CI"
      ],
      "os": [
        "Linux",
        "macOS",
        "Windows"
      ],
      "arch": [
        "x64",
        "arm64"
      ],
      "isolation": "Depends on the executor: Docker and Kubernetes run jobs in containers, the Docker Autoscaler and Instance executors create cloud instances through fleeting plugins, and the Shell and SSH executors run on the host without a clean build environment",
      "isolation_short": "Depends on executor",
      "isolation_class": "mixed",
      "ephemeral": null,
      "open_source": true,
      "license": "MIT",
      "self_host": true,
      "byoc": true,
      "clouds": [
        "AWS",
        "Google Cloud",
        "Azure",
        "Kubernetes",
        "DigitalOcean",
        "VMware vSphere",
        "Bare metal"
      ],
      "gpu": true,
      "gpu_detail": "GPU use is documented for the Shell, Docker (gpus option with the NVIDIA Container Toolkit), Kubernetes and Docker Machine executors.",
      "hardware": null,
      "performance": null,
      "sizes": null,
      "max_size_short": null,
      "cache": "The docs describe a distributed cache that uses an object storage server (the example uses S3) to share the cache between Docker hosts, and a Docker registry mirror for autoscaled hosts.",
      "docker_layer_cache": null,
      "concurrency": "Set by the concurrent setting in config.toml; for autoscaling the docs set it to max_instances multiplied by capacity_per_instance",
      "job_limits": null,
      "static_ips": null,
      "ssh_access": null,
      "observability": "Embedded Prometheus metrics HTTP server",
      "setup": "Install the runner from a package, binary, Docker image, Helm chart or Operator, register it with your GitLab instance and choose an executor in config.toml. For autoscaling, add a fleeting plugin and run gitlab-runner fleeting install.",
      "runs_on": null,
      "pricing": "Free and open source; you pay for the infrastructure it runs on. The docs list it for the Free, Premium and Ultimate tiers.",
      "price_short": null,
      "free_tier": null,
      "compliance": null,
      "regions": null,
      "best_for": "Teams on GitLab CI/CD that want to run jobs on their own machines, clusters or autoscaled cloud instances",
      "overview": [
        "**GitLab Runner** is the application that works with GitLab CI/CD to run the jobs in a pipeline. It is written in Go, distributed as a single binary under the MIT licence, and runs on Linux, macOS and Windows. An executor decides where each job runs: Docker, Kubernetes, Shell, SSH, VirtualBox, Parallels, Custom, or the autoscaling Docker Autoscaler and Instance executors. It works with GitLab.com, GitLab Self-Managed and GitLab Dedicated.",
        "Autoscaling on public clouds uses the Docker Autoscaler and Instance executors, which create instances on demand through **fleeting** plugins. GitLab maintains plugins for AWS Auto Scaling groups, Google Cloud instance groups and Azure scale sets; community plugins cover VMware vSphere, DigitalOcean, Magalu Cloud and libvirt. With capacity and use count set to one, each job gets an instance that is deleted afterwards. The older Docker Machine executor is deprecated in favour of this autoscaler."
      ],
      "notable": [
        "GitLab Runner v19.6.0 was released on 19 November 2026",
        "The Docker Machine executor was deprecated in GitLab 17.5 and is scheduled for removal in GitLab 20.0 (May 2027)",
        "The Docker Autoscaler and Instance executors became generally available in GitLab Runner 17.1"
      ],
      "pros": [
        "You use GitLab CI/CD and want to run jobs on infrastructure you control",
        "You want a choice of executors, from containers on Docker or Kubernetes to full cloud instances",
        "You want autoscaling on AWS, Google Cloud or Azure with plugins that GitLab maintains",
        "You need runners on Linux, macOS and Windows from the same project",
        "You want one single-use instance per job, which the autoscaling executors document"
      ],
      "cons": [
        "The Docker Machine executor is deprecated and scheduled for removal in GitLab 20.0",
        "SSH, Shell, Parallels, VirtualBox, Custom and Docker Machine executors receive critical security updates only, with no new features planned",
        "Community fleeting plugins are maintained outside GitLab, which says it cannot test them",
        "When an instance runs several jobs, the Instance executor docs say there is little isolation between them and the jobs should be trusted"
      ],
      "verdict": "The official open-source runner for GitLab CI/CD, with executors ranging from a local shell to autoscaled cloud instances; it suits any team that self-hosts GitLab runners.",
      "use_cases": {
        "gitlab": "It is the official runner application for GitLab CI/CD.",
        "open-source": "The runner is MIT licensed and distributed as a single binary.",
        "byoc": "Fleeting plugins create job instances in your own AWS, Google Cloud or Azure account.",
        "gpu": "GPU configuration is documented for the Shell, Docker and Kubernetes executors.",
        "macos-ios": "The runner installs on macOS on Apple Silicon and Intel systems."
      },
      "peers": [
        "gitlab-hosted-runners",
        "actions-runner-controller",
        "garm",
        "buildkite-elastic-ci-stack",
        "terraform-aws-github-runner",
        "azure-pipelines-agents"
      ],
      "sources": [
        "https://docs.gitlab.com/runner/",
        "https://docs.gitlab.com/runner/executors/",
        "https://docs.gitlab.com/runner/executors/docker_autoscaler/",
        "https://docs.gitlab.com/runner/executors/instance/",
        "https://docs.gitlab.com/runner/executors/docker_machine/",
        "https://docs.gitlab.com/runner/fleet_scaling/fleeting/",
        "https://docs.gitlab.com/runner/runner_autoscale/",
        "https://docs.gitlab.com/runner/configuration/autoscale/",
        "https://docs.gitlab.com/runner/configuration/gpus/",
        "https://docs.gitlab.com/runner/install/",
        "https://docs.gitlab.com/runner/install/linux-manually/",
        "https://gitlab.com/gitlab-org/gitlab-runner",
        "https://gitlab.com/gitlab-org/gitlab-runner/-/releases",
        "https://gitlab.com/gitlab-org/gitlab-runner/-/blob/main/LICENSE"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "fireactions",
      "name": "Fireactions",
      "company": "Hostinger",
      "hq": {
        "city": "Vilnius",
        "region": null,
        "country": "LT",
        "source": "https://www.hostinger.com/about"
      },
      "kind": "open-source",
      "category": "open-source-autoscalers",
      "website": "https://fireactions.io",
      "docs": "https://fireactions.io/latest/",
      "pricing_url": null,
      "github": "https://github.com/hostinger/fireactions",
      "tagline": "Orchestrator that runs self-hosted GitHub Actions runners in ephemeral Firecracker microVMs on your own Linux hosts.",
      "ci_systems": [
        "GitHub Actions"
      ],
      "os": [
        "Linux"
      ],
      "arch": [
        "x64",
        "arm64"
      ],
      "isolation": "Each runner is a Firecracker microVM that is created from scratch and destroyed after the job, with no state kept between jobs",
      "isolation_short": "Firecracker microVM",
      "isolation_class": "microvm",
      "ephemeral": true,
      "open_source": true,
      "license": "Apache-2.0",
      "self_host": true,
      "byoc": true,
      "clouds": [
        "Bare metal"
      ],
      "gpu": false,
      "gpu_detail": "The FAQ states that Fireactions does not support GPU workloads, because Firecracker has no GPU support.",
      "hardware": "Any Linux host with KVM on x86_64 or aarch64: a bare-metal server or a VM with nested virtualization, plus a dedicated block device for containerd image storage",
      "performance": "The Fireactions docs cite a GitHub runner startup time of about 20 seconds (vendor claim)",
      "sizes": "vCPU count and memory are set per pool in the Firecracker machine configuration",
      "max_size_short": null,
      "cache": "A tutorial describes running a Docker registry mirror (pull-through cache) on the host for Docker Hub images.",
      "docker_layer_cache": null,
      "concurrency": "Each pool keeps a configured number of runner replicas",
      "job_limits": null,
      "static_ips": null,
      "ssh_access": true,
      "observability": "Prometheus metrics endpoint, with an example Grafana dashboard in the docs",
      "setup": "Install Firecracker, containerd, the CNI plugins and the Fireactions binary on a KVM-capable Linux host. Create a GitHub App for your organisation and define pools in a YAML configuration file.",
      "runs_on": "[self-hosted, fireactions]",
      "pricing": "Free and open source; you pay for the infrastructure it runs on.",
      "price_short": null,
      "free_tier": null,
      "compliance": null,
      "regions": null,
      "best_for": "Teams with their own Linux servers that want microVM isolation for GitHub Actions jobs without running Kubernetes",
      "overview": [
        "**Fireactions** is an orchestrator for GitHub Actions runners that Hostinger built for its own infrastructure and describes as BYOM, bring your own metal. A server process reads pools from a YAML file and keeps the configured number of runners available in each. Every runner is a [Firecracker](https://firecracker-microvm.github.io/) microVM booted from an OCI image, registered with a just-in-time token, and shut down when its job finishes.",
        "It is a standalone application, which the FAQ contrasts with Kubernetes-based Actions Runner Controller, and it runs on a bare-metal server or a VM with nested virtualization. Hostinger publishes Ubuntu 20.04, 22.04 and 24.04 base images with Docker preinstalled, and custom images can be built with Docker. The CLI can list machines, stream agent logs and open an SSH session to a running VM. Hostinger is the only adopter listed in the repository."
      ],
      "notable": [
        "v2.2.2 was released on 17 November 2026",
        "The README notes many breaking changes and names v2.0.0 as the stable version to use in production",
        "The docs say the project is in the early stages of development and that Hostinger uses it fully in production"
      ],
      "pros": [
        "You want each job in its own Firecracker microVM that is destroyed afterwards",
        "You have bare-metal Linux servers, or VMs with nested virtualization, on x86_64 or arm64",
        "You prefer a standalone binary configured with one YAML file to a Kubernetes operator",
        "You want to build runner images with ordinary Docker tooling",
        "You want to SSH into a running runner VM from the CLI when debugging"
      ],
      "cons": [
        "Only GitHub organisation accounts are supported",
        "GPU workloads are not supported",
        "The docs describe the project as still in the early stages of development",
        "Running it inside a VM rather than on bare metal may affect performance, according to the FAQ"
      ],
      "verdict": "An open-source Firecracker orchestrator from Hostinger for ephemeral GitHub Actions runners on your own Linux hosts; it suits teams that want microVM isolation without Kubernetes.",
      "use_cases": {
        "open-source": "Fireactions is Apache-2.0 licensed and runs as a standalone binary on hosts you operate.",
        "security": "Each job runs in its own Firecracker microVM, which the docs recommend for untrusted workloads such as CI jobs.",
        "byoc": "It runs on your own bare-metal servers or nested-virtualization VMs.",
        "arm64": "aarch64 hosts are supported alongside x86_64."
      },
      "peers": [
        "actuated",
        "actions-runner-controller",
        "garm",
        "myshoes",
        "github-hetzner-runners",
        "ubicloud"
      ],
      "sources": [
        "https://github.com/hostinger/fireactions",
        "https://github.com/hostinger/fireactions/releases",
        "https://github.com/hostinger/fireactions/blob/main/LICENSE",
        "https://github.com/hostinger/fireactions/blob/main/ADOPTERS.md",
        "https://fireactions.io/latest/",
        "https://github.com/hostinger/fireactions/blob/main/docs/user-guide/overview.md",
        "https://github.com/hostinger/fireactions/blob/main/docs/user-guide/installation.md",
        "https://github.com/hostinger/fireactions/blob/main/docs/user-guide/images.md",
        "https://github.com/hostinger/fireactions/blob/main/docs/help/faq.md",
        "https://www.hostinger.com/about"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "myshoes",
      "name": "myshoes",
      "company": "Community project (whywaita)",
      "hq": null,
      "kind": "open-source",
      "category": "open-source-autoscalers",
      "website": "https://github.com/whywaita/myshoes",
      "docs": "https://github.com/whywaita/myshoes/tree/master/docs",
      "pricing_url": null,
      "github": "https://github.com/whywaita/myshoes",
      "tagline": "Daemon that provisions self-hosted GitHub Actions runners on demand through pluggable providers such as LXD, OpenStack and AWS.",
      "ci_systems": [
        "GitHub Actions"
      ],
      "os": [
        "Linux"
      ],
      "arch": null,
      "isolation": "Runners are instances created by a provider plugin (LXD, OpenStack, AWS EC2 or your own), so the boundary depends on the provider; whether each instance is single-use is not stated in the README or setup docs",
      "isolation_short": "Depends on provider",
      "isolation_class": "mixed",
      "ephemeral": null,
      "open_source": true,
      "license": "MIT",
      "self_host": true,
      "byoc": true,
      "clouds": [
        "LXD",
        "OpenStack",
        "AWS"
      ],
      "gpu": null,
      "gpu_detail": "The README lists special hardware such as GPUs and FPGAs as a reason to use your own provider, without setup details.",
      "hardware": null,
      "performance": null,
      "sizes": "Nine named resource types from nano to 4xlarge, which each provider maps to its own machine specs",
      "max_size_short": null,
      "cache": null,
      "docker_layer_cache": null,
      "concurrency": "MAX_CONNECTIONS_TO_BACKEND caps connections to the provider (default 50)",
      "job_limits": null,
      "static_ips": null,
      "ssh_access": null,
      "observability": "Optional Docker Hub rate-limit metrics",
      "setup": "An admin runs the myshoes daemon with a MySQL database, a GitHub App and a provider plugin. Users install the GitHub App and register a repository or organization as a target through the REST API.",
      "runs_on": null,
      "pricing": "Free and open source; you pay for the infrastructure it runs on.",
      "price_short": null,
      "free_tier": null,
      "compliance": null,
      "regions": null,
      "best_for": "Platform teams that run a private cloud such as OpenStack or LXD and want to offer on-demand GitHub Actions runners to their organisation",
      "overview": [
        "**myshoes** is a Go daemon that provides auto-scaling self-hosted runners for GitHub Actions. It receives `workflow_job` webhooks through a GitHub App and asks a provider plugin, which the project calls a shoes provider, to add or delete a runner instance. Providers are separate binaries that speak gRPC; the project links providers for LXD, OpenStack and AWS EC2, and documents how to write your own.",
        "An administrator runs the daemon with a MySQL database and a machine image for the runners. Users then register a repository or an organization as a target through a REST API and pick a resource type from nano to 4xlarge; a repository can override the size set for its organization. GitHub Enterprise Server is supported through a `GITHUB_URL` setting. The README links talks about its use for an internal hosted-runner service at CyberAgent."
      ],
      "notable": [
        "v1.20.4 was released on 5 November 2026",
        "The README links a talk titled “Development myshoes and Provide Cycloud-hosted runner” and one on an OSS CI/CD platform at CyberAgent"
      ],
      "pros": [
        "You run private infrastructure such as LXD or OpenStack and want runners created on it",
        "You want to set a default runner size per organization and override it for individual repositories",
        "You use GitHub Enterprise Server",
        "You want to write a provider for your own platform as a gRPC plugin"
      ],
      "cons": [
        "A MySQL database is required",
        "GitHub (or your GitHub Enterprise Server) must be able to reach the myshoes webhook endpoint",
        "The README lists the GCP provider as not yet available",
        "Admins must prepare the runner machine image themselves; the setup guide strongly recommends preinstalling Docker and the runner package"
      ],
      "verdict": "A community-maintained runner provisioner with a plugin model for private clouds; it suits platform teams offering GitHub Actions runners on LXD, OpenStack or AWS.",
      "use_cases": {
        "open-source": "myshoes is MIT licensed and self-hosted, with providers as separate open-source plugins.",
        "byoc": "Providers create runners on your own LXD hosts, OpenStack cloud or AWS account."
      },
      "peers": [
        "garm",
        "actions-runner-controller",
        "terraform-aws-github-runner",
        "fireactions",
        "github-hetzner-runners",
        "cirun"
      ],
      "sources": [
        "https://github.com/whywaita/myshoes",
        "https://github.com/whywaita/myshoes/releases",
        "https://github.com/whywaita/myshoes/blob/master/LICENSE",
        "https://github.com/whywaita/myshoes/blob/master/docs/01_01_for_admin_setup.md",
        "https://github.com/whywaita/myshoes/blob/master/docs/02_01_for_user_setup.md",
        "https://github.com/whywaita/myshoes/blob/master/docs/03_how-to-develop-shoes.md",
        "https://github.com/whywaita/myshoes-providers"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "github-hetzner-runners",
      "name": "TestFlows GitHub Hetzner Runners",
      "company": "Katteli Inc. (TestFlows)",
      "hq": null,
      "kind": "open-source",
      "category": "open-source-autoscalers",
      "website": "https://github.com/testflows/TestFlows-GitHub-Hetzner-Runners",
      "docs": "https://github.com/testflows/TestFlows-GitHub-Hetzner-Runners/wiki",
      "pricing_url": null,
      "github": "https://github.com/testflows/TestFlows-GitHub-Hetzner-Runners",
      "tagline": "Self-contained Python service that creates a Hetzner Cloud server for each queued GitHub Actions job, with no webhooks or GitHub App.",
      "ci_systems": [
        "GitHub Actions"
      ],
      "os": [
        "Linux"
      ],
      "arch": [
        "x64",
        "arm64"
      ],
      "isolation": "Each job gets a dedicated Hetzner Cloud VM that is not shared between jobs; with recycling (on by default) a powered-off server is rebuilt from its image before reuse, unless the no-rebuild mode for trusted single-tenant setups is chosen",
      "isolation_short": "Hetzner Cloud VM",
      "isolation_class": "vm",
      "ephemeral": true,
      "open_source": true,
      "license": "Apache-2.0",
      "self_host": true,
      "byoc": true,
      "clouds": [
        "Hetzner"
      ],
      "gpu": null,
      "gpu_detail": null,
      "hardware": "Any Hetzner Cloud server type, image and location, selected with job labels",
      "performance": null,
      "sizes": null,
      "max_size_short": null,
      "cache": "Up to sixteen persistent Hetzner Cloud volumes of 10 GB to 10 TB per runner, requested with volume labels and mounted under /mnt. The wiki describes ways to keep Docker data on them and warns that volumes are never deleted automatically.",
      "docker_layer_cache": true,
      "concurrency": "A maximum number of runners can be set, as can a limit per workflow run",
      "job_limits": null,
      "static_ips": null,
      "ssh_access": true,
      "observability": "Prometheus metrics endpoint and an embedded dashboard showing cost per hour, servers, runners, queued and running jobs, scale-up errors and logs",
      "setup": "pip install testflows.github.hetzner.runners, then supply a classic GitHub token, the repository and a Hetzner Cloud token. Run the service locally or deploy it to a Hetzner instance with github-hetzner-runners cloud deploy.",
      "runs_on": "[self-hosted, type-cpx22]",
      "pricing": "Free and open source; you pay for the infrastructure it runs on. The README notes that Hetzner Cloud bills servers hourly, so without recycling the minimum cost of a job is one server-hour plus a public IPv4 address.",
      "price_short": null,
      "free_tier": null,
      "compliance": null,
      "regions": null,
      "best_for": "Single-repository projects that want low-cost x64 or arm64 GitHub Actions runners on Hetzner Cloud with very little setup",
      "overview": [
        "**TestFlows GitHub Hetzner Runners** is a Python command-line service that watches a repository for queued GitHub Actions jobs and creates a Hetzner Cloud server for each one. The server registers as an ephemeral runner, powers off when the job completes, and is then deleted or recycled. It needs only a classic GitHub token and a Hetzner Cloud token: there are no webhooks, no GitHub App and no Kubernetes.",
        "Job labels select the server type, image, location and network, and both x64 and arm64 servers are supported. Because Hetzner bills by the hour, powered-off servers are recycled by default and rebuilt from their image for the next job. Standby pools, persistent caching volumes, cost estimation, Prometheus metrics and an embedded dashboard are included. TestFlows presents a figure of up to 33× cheaper than GitHub-hosted runners as an upper bound (vendor claim). The README lists Altinity’s ClickHouse builds among its users."
      ],
      "notable": [
        "Release 1.10.260921.1182403 was published on 21 September 2026",
        "Persistent caching volumes were added in version 1.9, according to the README",
        "The README lists Altinity’s ClickHouse builds and regression tests as users"
      ],
      "pros": [
        "You want each job on its own Hetzner Cloud VM with full root access and native Docker",
        "You want a setup with no webhooks, no GitHub App and no Kubernetes",
        "You need arm64 as well as x64 runners",
        "You want persistent volumes for caches that survive between jobs",
        "You want to SSH into a runner server with a single command when debugging"
      ],
      "cons": [
        "Group runners are not supported",
        "A unique Hetzner Cloud project must be used for each repository",
        "A classic GitHub token is required; fine-grained tokens are not supported",
        "The README says manual monitoring is required to make sure servers are cleaned up and costs stay under control"
      ],
      "verdict": "A single-command autoscaler that puts each GitHub Actions job on its own Hetzner Cloud server; it suits individual repositories that want inexpensive runners and accept hands-on operation.",
      "use_cases": {
        "lower-cost": "TestFlows presents Hetzner Cloud servers as up to 33× cheaper than GitHub-hosted runners at full hourly utilisation (vendor claim).",
        "arm64": "Hetzner arm64 server types are supported alongside x64.",
        "open-source": "The service is Apache-2.0 licensed and installed from PyPI.",
        "byoc": "Servers are created in your own Hetzner Cloud project.",
        "docker-builds": "Runners are full VMs with Docker preinstalled, and caching volumes can hold Docker data between jobs."
      },
      "peers": [
        "terraform-aws-github-runner",
        "garm",
        "cirun",
        "fireactions",
        "actions-runner-controller",
        "ubicloud",
        "runs-on"
      ],
      "sources": [
        "https://github.com/testflows/TestFlows-GitHub-Hetzner-Runners",
        "https://github.com/testflows/TestFlows-GitHub-Hetzner-Runners/releases",
        "https://github.com/testflows/TestFlows-GitHub-Hetzner-Runners/blob/main/LICENSE",
        "https://github.com/testflows/TestFlows-GitHub-Hetzner-Runners/wiki/Recycling-Powered-Off-Servers",
        "https://github.com/testflows/TestFlows-GitHub-Hetzner-Runners/wiki/Using-Caching-Volumes",
        "https://github.com/testflows/TestFlows-GitHub-Hetzner-Runners/wiki/Opening-The-SSH-Client-To-The-Server",
        "https://testflows.com/"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "buildkite-elastic-ci-stack",
      "name": "Elastic CI Stack for AWS",
      "company": "Buildkite Pty Ltd",
      "hq": {
        "city": "Melbourne",
        "region": "Victoria",
        "country": "AU",
        "source": "https://buildkite.com/about/legal/privacy-policy/"
      },
      "kind": "open-source",
      "category": "open-source-autoscalers",
      "website": "https://buildkite.com/docs/agent/self-hosted/aws/elastic-ci-stack",
      "docs": "https://buildkite.com/docs/agent/self-hosted/aws/elastic-ci-stack",
      "pricing_url": null,
      "github": "https://github.com/buildkite/elastic-ci-stack-for-aws",
      "tagline": "CloudFormation template (with a Terraform option) that runs an autoscaling group of Buildkite agents on EC2 in your own AWS account.",
      "ci_systems": [
        "Buildkite"
      ],
      "os": [
        "Linux",
        "Windows"
      ],
      "arch": [
        "x64",
        "arm64"
      ],
      "isolation": "Agents run on EC2 instances in an Auto Scaling group inside your VPC; instances are reused between jobs by default, and a template parameter can terminate the instance after each job",
      "isolation_short": "EC2 instance",
      "isolation_class": "vm",
      "ephemeral": null,
      "open_source": true,
      "license": "MIT",
      "self_host": true,
      "byoc": true,
      "clouds": [
        "AWS"
      ],
      "gpu": null,
      "gpu_detail": null,
      "hardware": "EC2 instance types set with the InstanceTypes parameter (default t3.large); local NVMe instance storage can be enabled on Linux",
      "performance": null,
      "sizes": null,
      "max_size_short": null,
      "cache": "Git mirrors can be enabled as an agent experiment, and on Linux the builds directory can be placed on local NVMe instance storage.",
      "docker_layer_cache": null,
      "concurrency": "Template defaults are MinSize 0 and MaxSize 10 instances with one agent per instance; all three are configurable",
      "job_limits": null,
      "static_ips": null,
      "ssh_access": true,
      "observability": "CloudWatch Logs for system and Buildkite agent events, and CloudWatch metrics from the Buildkite API",
      "setup": "Launch the CloudFormation template (or the Terraform module) with a Buildkite agent token, then set parameters such as instance types and the minimum and maximum number of instances.",
      "runs_on": null,
      "pricing": "Free and open source; you pay for the infrastructure it runs on. Buildkite notes that the AWS services are pay-as-you-go and that Buildkite services are billed according to your plan.",
      "price_short": null,
      "free_tier": null,
      "compliance": null,
      "regions": "All AWS Regions except China and US GovCloud",
      "best_for": "Buildkite customers on AWS that want an autoscaling agent fleet in their own VPC with a vendor-maintained template",
      "overview": [
        "The **Elastic CI Stack for AWS** is Buildkite’s template for a private, autoscaling cluster of Buildkite agents. Each deployment contains an EC2 Auto Scaling group and a launch template, and a Lambda function adjusts the group according to build activity. It is deployed with CloudFormation, and Buildkite also publishes a Terraform module. Docker and Docker Compose are preinstalled, and secrets can be delivered per pipeline from an S3 bucket.",
        "Buildkite builds AMIs for Amazon Linux 2023 and Ubuntu 24.04 on x86 and Arm, and for Windows Server 2022 and 2025 on x86. The minimum size defaults to zero and instances are reused until idle unless you set them to terminate after each job. Spot instances, scheduled scaling and several agents per instance are configurable. Instances can be reached over SSH on Linux or through AWS Systems Manager. A separate CloudFormation setup covers EC2 Mac."
      ],
      "notable": [
        "v7.13.0 was released on 19 November 2026",
        "The README states that new features are developed only for v7 and that security fixes continue to be backported to v6",
        "Buildkite says it aims to publish a tagged release at the end of each month when the main branch has changed"
      ],
      "pros": [
        "You use Buildkite and want agents inside your own AWS account and VPC",
        "You want a template maintained by Buildkite, with prebuilt Linux and Windows AMIs",
        "You want the fleet to scale with build activity and down to a minimum of zero instances",
        "You need Linux on x86 and Arm as well as Windows Server agents",
        "You want to use spot instances and scheduled scaling to manage EC2 cost"
      ],
      "cons": [
        "The README states that the repository has not been reviewed by security researchers",
        "Builds can reach the EC2 metadata server and act with the IAM permissions of the instance",
        "China and US GovCloud regions are not supported",
        "SSH access, instance NVMe storage and Git LFS are listed as Linux-only features"
      ],
      "verdict": "Buildkite’s own open-source template for autoscaling agents on EC2; it suits Buildkite customers that keep builds inside their AWS account.",
      "use_cases": {
        "byoc": "The stack creates the agents, Auto Scaling group and VPC in your own AWS account.",
        "open-source": "The template and AMI build scripts are MIT licensed.",
        "windows": "Buildkite publishes Windows Server 2022 and 2025 AMIs for the stack.",
        "arm64": "Amazon Linux 2023 and Ubuntu 24.04 AMIs are published for 64-bit Arm."
      },
      "peers": [
        "buildkite-hosted-agents",
        "gitlab-runner",
        "terraform-aws-github-runner",
        "actions-runner-controller",
        "runs-on",
        "aws-codebuild-runners"
      ],
      "sources": [
        "https://github.com/buildkite/elastic-ci-stack-for-aws",
        "https://github.com/buildkite/elastic-ci-stack-for-aws/releases",
        "https://github.com/buildkite/elastic-ci-stack-for-aws/blob/main/Licence.md",
        "https://github.com/buildkite/elastic-ci-stack-for-aws/blob/main/templates/aws-stack.yml",
        "https://buildkite.com/docs/agent/self-hosted/aws/elastic-ci-stack",
        "https://buildkite.com/docs/agent/self-hosted/aws/elastic-ci-stack/ec2-linux-and-windows",
        "https://buildkite.com/about/legal/privacy-policy/"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "ec2-github-runner",
      "name": "ec2-github-runner",
      "company": "Community project (machulav)",
      "hq": null,
      "kind": "open-source",
      "category": "open-source-autoscalers",
      "website": "https://github.com/machulav/ec2-github-runner",
      "docs": "https://github.com/machulav/ec2-github-runner#usage",
      "pricing_url": null,
      "github": "https://github.com/machulav/ec2-github-runner",
      "tagline": "GitHub Action that starts an EC2 instance as a self-hosted runner at the start of a workflow and terminates it at the end.",
      "ci_systems": [
        "GitHub Actions"
      ],
      "os": [
        "Linux"
      ],
      "arch": null,
      "isolation": "Each workflow run starts its own EC2 instance in your VPC and a stop job terminates it; with the optional JIT mode the runner is single-use and deregisters after one job",
      "isolation_short": "EC2 instance",
      "isolation_class": "vm",
      "ephemeral": null,
      "open_source": true,
      "license": "MIT",
      "self_host": true,
      "byoc": true,
      "clouds": [
        "AWS"
      ],
      "gpu": null,
      "gpu_detail": null,
      "hardware": "Any EC2 instance type, on-demand or spot, from an AMI you provide",
      "performance": null,
      "sizes": null,
      "max_size_short": null,
      "cache": null,
      "docker_layer_cache": null,
      "concurrency": null,
      "job_limits": null,
      "static_ips": null,
      "ssh_access": null,
      "observability": "An optional debug mode streams the EC2 serial console output into the GitHub Actions log during runner startup",
      "setup": "Add a start job and a stop job that call machulav/ec2-github-runner@v2 with AWS credentials, a GitHub personal access token, an AMI, a subnet and a security group. The job in between uses the label output as its runs-on value.",
      "runs_on": "${{ needs.start-runner.outputs.label }}",
      "pricing": "Free and open source; you pay for the infrastructure it runs on.",
      "price_short": null,
      "free_tier": null,
      "compliance": null,
      "regions": null,
      "best_for": "Repositories that occasionally need a job to run on a specific EC2 instance type or inside a private VPC, without operating a standing autoscaler",
      "overview": [
        "**ec2-github-runner** is a GitHub Action rather than a standing service. In the documented workflow, a first job on a GitHub-hosted runner calls the action in `start` mode to launch an EC2 instance and register it as a self-hosted runner with a unique label. The main job runs on that label. A final job calls the action in `stop` mode to terminate the instance and remove the runner. Everything is driven from the workflow file.",
        "The README gives three uses: reaching private resources in a VPC, picking any EC2 instance type, and saving cost. Inputs cover the AMI, subnet, security group, IAM role, spot market type, block devices and a pre-runner script. A JIT option registers a single-use runner, and a multi-AZ option tries several subnets or regions in turn. Runners are registered at repository level with a personal access token."
      ],
      "notable": [
        "v2.6.1 was released on 10 April 2026",
        "The README documents JIT runners, multi-AZ failover and a debug mode as advanced options"
      ],
      "pros": [
        "You want an EC2 runner that exists only for the duration of one workflow",
        "You need a job to reach private resources in your VPC",
        "You want to pick the EC2 instance type per workflow, on-demand or spot",
        "You do not want to deploy and operate webhooks, Lambda functions or a controller"
      ],
      "cons": [
        "The workflow must include a stop job; otherwise the EC2 instance keeps running after the workflow finishes",
        "It needs a GitHub personal access token with the repo scope and manages runners at repository level",
        "You must prepare an AMI with Docker and Git installed, or install them at boot",
        "The README repeats GitHub’s recommendation not to use self-hosted runners with public repositories"
      ],
      "verdict": "A GitHub Action that launches and removes one EC2 runner per workflow run; it suits repositories with occasional needs for custom hardware or VPC access rather than a shared runner fleet.",
      "use_cases": {
        "open-source": "The action is MIT licensed and runs from your own workflow.",
        "byoc": "The runner is an EC2 instance in a subnet and security group that you choose.",
        "lower-cost": "The README describes running jobs on cheaper EC2 instances, including spot, as one of its uses."
      },
      "peers": [
        "terraform-aws-github-runner",
        "runs-on",
        "aws-codebuild-runners",
        "cirun",
        "actions-runner-controller",
        "hyperenv"
      ],
      "sources": [
        "https://github.com/machulav/ec2-github-runner",
        "https://github.com/machulav/ec2-github-runner/releases",
        "https://github.com/machulav/ec2-github-runner/blob/main/LICENSE"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "terraform-aws-gitlab-runner",
      "name": "terraform-aws-gitlab-runner",
      "company": "cattle-ops (community project)",
      "hq": null,
      "kind": "open-source",
      "category": "open-source-autoscalers",
      "website": "https://registry.terraform.io/modules/cattle-ops/gitlab-runner/aws",
      "docs": "https://github.com/cattle-ops/terraform-aws-gitlab-runner/blob/main/docs/usage.md",
      "pricing_url": null,
      "github": "https://github.com/cattle-ops/terraform-aws-gitlab-runner",
      "tagline": "Terraform module that deploys GitLab Runner on AWS and autoscales its job workers on EC2 spot instances.",
      "ci_systems": [
        "GitLab CI"
      ],
      "os": [
        "Linux"
      ],
      "arch": null,
      "isolation": "A runner agent on one EC2 instance creates worker EC2 instances with the docker+machine executor (the default) or the Docker Autoscaler executor; in the plain docker scenario builds run on the agent instance itself",
      "isolation_short": "EC2 instance",
      "isolation_class": "vm",
      "ephemeral": null,
      "open_source": true,
      "license": "MIT",
      "self_host": true,
      "byoc": true,
      "clouds": [
        "AWS"
      ],
      "gpu": null,
      "gpu_detail": null,
      "hardware": "EC2 instances, spot by default; a spot fleet scenario spreads workers over several instance types and availability zones",
      "performance": null,
      "sizes": null,
      "max_size_short": null,
      "cache": "Creates an S3 bucket as a shared runner cache by default, with a lifecycle policy that removes old objects; the bucket can be managed outside the module to share it between runners.",
      "docker_layer_cache": null,
      "concurrency": null,
      "job_limits": null,
      "static_ips": null,
      "ssh_access": true,
      "observability": "Runner logs are streamed to CloudWatch",
      "setup": "Store a pre-registered GitLab runner token in an SSM parameter and apply the Terraform module with your VPC and subnet settings.",
      "runs_on": null,
      "pricing": "Free and open source; you pay for the infrastructure it runs on.",
      "price_short": null,
      "free_tier": null,
      "compliance": null,
      "regions": null,
      "best_for": "Teams on GitLab CI that manage AWS with Terraform and want autoscaling runners on spot instances",
      "overview": [
        "**terraform-aws-gitlab-runner** is a community Terraform module that creates a GitLab Runner deployment on AWS. A runner agent runs on one EC2 instance and starts worker instances for builds, on spot instances by default, using the `docker+machine` executor. The module registers the agent automatically, streams logs to CloudWatch, and creates an S3 bucket as a shared cache with lifecycle rules.",
        "The README describes three scenarios: one agent that autoscales workers, several agents created by instantiating the module more than once, and a plain Docker runner without autoscaling. A newer scenario uses GitLab’s Docker Autoscaler executor in place of Docker Machine and needs a worker AMI with Docker Engine preinstalled. Scheduled scaling, a Lambda function for graceful termination and Session Manager access to instances are included. The module is published in the Terraform Registry."
      ],
      "notable": [
        "Version 9.6.0 was released on 9 October 2026",
        "The README traces the module to a 2017 blog post on running GitLab runners on spot instances"
      ],
      "pros": [
        "You use GitLab CI and want runners on EC2 spot instances defined in Terraform",
        "You want an S3-backed shared cache created for you",
        "You want to run several runner agents with different configurations from the same module",
        "You want to move from Docker Machine to the Docker Autoscaler executor within the same module"
      ],
      "cons": [
        "The default docker+machine executor depends on Docker Machine, which the usage guide notes is no longer maintained by Docker",
        "The Docker Autoscaler scenario needs an AMI with Docker Engine preinstalled, which base Amazon Linux and Ubuntu AMIs do not have",
        "The plain docker scenario does not support auto-scaling",
        "Removing the module requires a specific procedure so that key pairs and spot requests are cleaned up"
      ],
      "verdict": "A community Terraform module for autoscaling GitLab Runner on AWS spot instances; it suits GitLab teams that already manage AWS with Terraform.",
      "use_cases": {
        "gitlab": "The module deploys and registers GitLab Runner agents and their autoscaled workers.",
        "open-source": "The module is MIT licensed and published in the Terraform Registry.",
        "byoc": "All instances, the cache bucket and the Lambda functions are created in your own AWS account.",
        "lower-cost": "Workers run on EC2 spot instances by default."
      },
      "peers": [
        "gitlab-runner",
        "gitlab-hosted-runners",
        "terraform-aws-github-runner",
        "buildkite-elastic-ci-stack",
        "aws-codebuild-runners"
      ],
      "sources": [
        "https://github.com/cattle-ops/terraform-aws-gitlab-runner",
        "https://github.com/cattle-ops/terraform-aws-gitlab-runner/releases",
        "https://github.com/cattle-ops/terraform-aws-gitlab-runner/blob/main/docs/usage.md",
        "https://github.com/cattle-ops/terraform-aws-gitlab-runner/blob/main/CHANGELOG.md",
        "https://github.com/cattle-ops/terraform-aws-gitlab-runner/blob/main/LICENSE"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "github-hosted-runners",
      "name": "GitHub-hosted runners",
      "company": "GitHub, Inc.",
      "hq": {
        "city": "San Francisco",
        "region": "California",
        "country": "US",
        "source": "https://docs.github.com/en/site-policy/privacy-policies/github-general-privacy-statement"
      },
      "kind": "platform",
      "category": "platform-hosted",
      "website": "https://github.com/features/actions",
      "docs": "https://docs.github.com/en/actions/concepts/runners/github-hosted-runners",
      "pricing_url": "https://docs.github.com/en/billing/reference/actions-runner-pricing",
      "github": "https://github.com/actions/runner-images",
      "tagline": "GitHub’s own Actions runners: a new VM per job on Linux, Windows and macOS, with larger, arm64 and GPU sizes billed per minute.",
      "ci_systems": [
        "GitHub Actions"
      ],
      "os": [
        "Linux",
        "macOS",
        "Windows"
      ],
      "arch": [
        "x64",
        "arm64"
      ],
      "isolation": "Each job runs in a new virtual machine hosted by GitHub; the 1-vCPU ubuntu-slim runner instead runs in an unprivileged container on a shared VM",
      "isolation_short": "Fresh VM",
      "isolation_class": "vm",
      "ephemeral": true,
      "open_source": false,
      "license": null,
      "self_host": false,
      "byoc": false,
      "clouds": null,
      "gpu": true,
      "gpu_detail": "GPU larger runners have 4 vCPU, 28 GB RAM and one Tesla T4 with 16 GB VRAM, at $0.052/min on Linux and $0.102/min on Windows.",
      "hardware": "Azure virtual machines, with eligible 2-core and 4-core Linux jobs also scheduled on AWS; standard macOS runners are 3-core M1 or 4-core Intel, and the macOS XLarge runner is a 5-core M2",
      "performance": null,
      "sizes": "Standard: 1 vCPU / 5 GB (ubuntu-slim); 2 vCPU / 8 GB in private and 4 vCPU / 16 GB in public repositories on Linux and Windows, x64 and arm64. Larger runners: 2 to 96 vCPU on x64 and 2 to 64 vCPU on arm64; macOS 3 to 12 cores",
      "max_size_short": "96 vCPU / 384 GB",
      "cache": "Dependency caching through the cache action, with 10 GB per repository included; entries not accessed for 7 days are removed and the oldest are evicted at the limit. The limit can be raised, billed at $0.07 per GB per month. Larger runners can use custom VM images.",
      "docker_layer_cache": null,
      "concurrency": "Standard runners: 20 concurrent jobs on Free, 40 on Pro, 60 on Team and 500 on Enterprise, of which 5 macOS (50 on Enterprise). Larger runners: up to 1,000 concurrent jobs and 100 GPU jobs",
      "job_limits": "6 hours of execution per job (15 minutes on ubuntu-slim); 35 days per workflow run; 256 jobs per matrix",
      "static_ips": true,
      "ssh_access": null,
      "observability": "Actions usage metrics and performance metrics (job run time, queue time, failure rate) for organizations and repositories.",
      "setup": "Nothing to install: set runs-on to a GitHub-hosted label. Larger runners are created by an organization or enterprise owner on GitHub Team or Enterprise Cloud and need a payment method on file.",
      "runs_on": "ubuntu-latest",
      "pricing": "Private repositories, per minute: Linux 2-core x64 $0.006, Linux 2-core arm64 $0.005, Linux 1-core $0.002, Windows 2-core $0.010, macOS 3- or 4-core $0.062. Larger Linux x64 runners run from $0.012 (4-core) to $0.252 (96-core); larger macOS runners are $0.077 (12-core Intel) and $0.102 (M2 Pro). Usage is rounded up to the whole minute.",
      "price_short": "$0.006/min",
      "free_tier": "Standard runners are free in public repositories. Private repositories get $12 of included compute usage per month on GitHub Free, $18 on Pro and Team, and $300 on Enterprise Cloud",
      "compliance": null,
      "regions": "Regions are not documented for github.com. GitHub says Windows runners and other Linux runners run on Microsoft Azure, eligible Linux jobs on Azure or AWS, and macOS runners in Azure data centers",
      "best_for": "Teams on GitHub Actions that want runners with no setup, and public repositories that can use the free standard runners",
      "overview": [
        "**GitHub-hosted runners** are the machines GitHub provides for GitHub Actions jobs. A workflow selects one with a `runs-on` label such as `ubuntu-latest`, and GitHub starts a new virtual machine for the job from an image it maintains in the [runner-images repository](https://github.com/actions/runner-images). Standard runners cover Linux, Windows and macOS on x64 and arm64. They are free in public repositories and draw on a monthly included amount in private ones.",
        "Organizations on GitHub Team or Enterprise Cloud can also create **larger runners**: 2 to 96 vCPU machines, GPU machines, static IP ranges, Azure private networking and custom images, all billed per minute, including in public repositories. A 1-vCPU `ubuntu-slim` runner, generally available since January 2026, runs short automation jobs in a container. Since October 2026 eligible Linux jobs may run on AWS as well as Azure."
      ],
      "notable": [
        "GitHub reduced GitHub-hosted runner prices by up to 39% on January 1, 2026; a $0.002 per-minute charge for self-hosted runners announced on December 16, 2025 was postponed",
        "Linux and Windows arm64 standard runners became available in private repositories on January 29, 2026",
        "The 1-vCPU ubuntu-slim runner became generally available on January 22, 2026",
        "On October 28, 2026 GitHub began scheduling eligible Linux jobs across a multi-cloud pool that includes AWS"
      ],
      "pros": [
        "You want runners that need no installation or account beyond GitHub itself",
        "You maintain public repositories, where standard runners are free",
        "You need Linux, Windows and macOS, on x64 and arm64, from one provider",
        "You want every job to start in a clean VM that is not reused",
        "You need GPU machines or static IP ranges without operating your own runners"
      ],
      "cons": [
        "Larger runners are billed in public repositories and need a GitHub Team or Enterprise Cloud plan",
        "Jobs are limited to 6 hours of execution time",
        "Standard runners have 14 GB of SSD storage and at most 4 vCPU",
        "Static IP addresses and Azure private networking are not available for macOS runners"
      ],
      "verdict": "The default runners for GitHub Actions, suited to teams that prefer zero setup and to public repositories, with larger sizes available per minute on paid plans.",
      "use_cases": {
        "arm64": "Linux and Windows arm64 standard runners are available in public and private repositories, with larger arm64 runners up to 64 vCPU.",
        "macos-ios": "macOS runners are offered on M1, M2 and Intel hardware with macOS 15 and 26 images and an Xcode 27 image.",
        "windows": "Windows Server 2022 and 2025 and Windows 11 arm64 images are available on standard and larger runners.",
        "gpu": "Larger runners include a 4-vCPU Tesla T4 machine on Linux and Windows.",
        "security": "GitHub documents its hosted runners as clean, ephemeral virtual machines and says self-hosted runners should almost never be used for public repositories."
      },
      "peers": [
        "blacksmith",
        "depot",
        "namespace",
        "warpbuild",
        "ubicloud",
        "runs-on",
        "actions-runner-controller",
        "gitlab-hosted-runners"
      ],
      "sources": [
        "https://docs.github.com/en/billing/reference/actions-runner-pricing",
        "https://docs.github.com/en/billing/concepts/product-billing/github-actions",
        "https://docs.github.com/en/actions/reference/runners/github-hosted-runners",
        "https://docs.github.com/en/actions/reference/runners/larger-runners",
        "https://docs.github.com/en/actions/concepts/runners/github-hosted-runners",
        "https://docs.github.com/en/actions/reference/limits",
        "https://docs.github.com/en/actions/reference/workflows-and-actions/dependency-caching",
        "https://docs.github.com/en/actions/reference/security/secure-use",
        "https://docs.github.com/en/actions/concepts/metrics",
        "https://github.blog/changelog/2025-12-16-coming-soon-simpler-pricing-and-a-better-experience-for-github-actions/",
        "https://github.blog/changelog/2026-01-01-reduced-pricing-for-github-hosted-runners-usage/",
        "https://github.blog/changelog/2026-01-22-1-vcpu-linux-runner-now-generally-available-in-github-actions/",
        "https://github.blog/changelog/2026-01-29-arm64-standard-runners-are-now-available-in-private-repositories/",
        "https://github.blog/changelog/2026-10-28-github-actions-adds-multi-cloud-compute-to-improve-reliability/",
        "https://docs.github.com/en/site-policy/privacy-policies/github-general-privacy-statement"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "gitlab-hosted-runners",
      "name": "GitLab-hosted runners",
      "company": "GitLab Inc.",
      "hq": null,
      "kind": "platform",
      "category": "platform-hosted",
      "website": "https://about.gitlab.com",
      "docs": "https://docs.gitlab.com/ci/runners/hosted_runners/",
      "pricing_url": "https://about.gitlab.com/pricing/",
      "github": null,
      "tagline": "GitLab.com’s built-in runners: each job gets a single-use VM, metered in compute minutes with a cost factor per machine size.",
      "ci_systems": [
        "GitLab CI"
      ],
      "os": [
        "Linux",
        "macOS",
        "Windows"
      ],
      "arch": [
        "x64",
        "arm64"
      ],
      "isolation": "Each job runs in a newly provisioned VM that serves a single job and is deleted right after it; firewall rules block inbound traffic and traffic between VMs",
      "isolation_short": "Ephemeral VM",
      "isolation_class": "vm",
      "ephemeral": true,
      "open_source": false,
      "license": null,
      "self_host": false,
      "byoc": false,
      "clouds": null,
      "gpu": true,
      "gpu_detail": "One GPU machine type on Linux x86-64 for Premium and Ultimate: 4 vCPU, 15 GB RAM and 1 NVIDIA Tesla T4 (or similar) with 16 GB, at cost factor 7.",
      "hardware": "Linux runners are Google Compute Engine VMs running Container-Optimized OS; macOS runners are Apple silicon M1 and M2 Pro machines. GitLab notes that machine and processor types might change",
      "performance": null,
      "sizes": "Linux x86-64: 2 vCPU / 8 GB to 32 vCPU / 128 GB; Linux Arm64: 2 vCPU / 8 GB to 8 vCPU / 32 GB; macOS: 4 vCPU / 8 GB (M1) and 6 vCPU / 16 GB (M2 Pro); Windows: 2 vCPU / 7.5 GB",
      "max_size_short": "32 vCPU / 128 GB",
      "cache": "Hosted runners share a distributed cache in Google Cloud Storage. Cache contents not updated in the last 14 days are removed, and an uploaded cache archive can be at most 5 GB.",
      "docker_layer_cache": null,
      "concurrency": "Maximum jobs in active pipelines on GitLab.com: 500 on Free, 20,000 on Premium and 60,000 on Ultimate",
      "job_limits": "Jobs on GitLab.com hosted runners time out after 3 hours, regardless of the timeout configured in the project",
      "static_ips": null,
      "ssh_access": null,
      "observability": "Compute minute usage per namespace on the instance runner usage dashboard, with notifications as a quota is approached.",
      "setup": "Enabled by default for all projects on GitLab.com. Untagged jobs run on the small Linux x86-64 runner; add a tags entry in .gitlab-ci.yml to pick another machine type.",
      "runs_on": "saas-linux-medium-amd64",
      "pricing": "No per-minute price is published. Jobs consume compute minutes at job duration times a cost factor: Linux small 1, medium 2, large 3, xlarge 6, 2xlarge 12; GPU 7; macOS M1 6 and M2 Pro 12 (beta); Windows 1 (beta). Additional compute minutes cost $10 per 1,000 minutes as a one-time purchase.",
      "price_short": "$10 per 1,000 compute minutes",
      "free_tier": "400 compute minutes per month on the Free tier; Premium includes 10,000 and Ultimate 50,000 per month",
      "compliance": null,
      "regions": "Linux x86-64 VMs run in Google Cloud us-east1 and us-east5; Linux Arm64 VMs run in us-central1",
      "best_for": "Teams on GitLab.com that want CI jobs to run with no runner setup and can work within the included compute minutes",
      "overview": [
        "**GitLab-hosted runners** are the runners GitLab operates for projects on GitLab.com. They are enabled by default: a job with no tags runs on the small Linux x86-64 machine, and a `tags` entry such as `saas-linux-medium-amd64` selects another size. Each job gets a newly provisioned VM that is deleted afterwards. Linux runners use the docker+machine executor on Google Compute Engine and run in privileged mode, so Docker-in-Docker builds work.",
        "Usage is metered in **compute minutes** rather than priced per minute: job duration is multiplied by a cost factor that rises with machine size, and each tier includes a monthly quota. Larger Linux machines, the GPU runner and macOS runners require Premium or Ultimate. GitLab’s documentation labels the macOS and Windows runners as beta. Separate hosted runners exist for GitLab Dedicated."
      ],
      "notable": [
        "As of November 2026 GitLab’s documentation labels hosted runners on macOS and on Windows as beta",
        "As of November 2026 the documented macOS images are macos-15-xcode-16 and macos-26-xcode-26, with macos-27-xcode-27 in beta",
        "GitLab states a service level objective that 90% of CI/CD jobs start executing in 120 seconds or less"
      ],
      "pros": [
        "You host code on GitLab.com and want pipelines to run without registering any runner",
        "You want each job in its own VM that is deleted after the job",
        "You need Docker-in-Docker, which the Linux runners support by running in privileged mode",
        "You need Linux Arm64 or a GPU machine alongside x86-64 from the same platform"
      ],
      "cons": [
        "Hosted runners on macOS and on Windows are in beta",
        "Jobs time out after 3 hours regardless of the project timeout",
        "Large and bigger Linux machines, the GPU runner and macOS runners are limited to Premium and Ultimate",
        "GitLab says macOS machine availability is limited and jobs might queue for extended periods"
      ],
      "verdict": "The built-in runners for GitLab.com, suited to teams that want no runner administration and whose usage fits the compute-minute quota of their tier.",
      "use_cases": {
        "gitlab": "These are GitLab’s own hosted runners for GitLab CI/CD on GitLab.com, enabled by default for all projects.",
        "arm64": "Linux Arm64 machine types with 2, 4 and 8 vCPU are documented.",
        "macos-ios": "macOS runners on M1 and M2 Pro hardware with Xcode images are available in beta for Premium and Ultimate.",
        "gpu": "A GPU-enabled Linux machine type with an NVIDIA Tesla T4 (or similar) is available on Premium and Ultimate.",
        "docker-builds": "Linux runners run in privileged mode to support Docker-in-Docker image builds."
      },
      "peers": [
        "gitlab-runner",
        "github-hosted-runners",
        "buildkite-hosted-agents",
        "azure-pipelines-agents",
        "aws-codebuild-runners",
        "cirun"
      ],
      "sources": [
        "https://docs.gitlab.com/ci/runners/hosted_runners/",
        "https://docs.gitlab.com/ci/runners/hosted_runners/linux/",
        "https://docs.gitlab.com/ci/runners/hosted_runners/macos/",
        "https://docs.gitlab.com/ci/runners/hosted_runners/windows/",
        "https://docs.gitlab.com/ci/runners/hosted_runners/gpu_enabled/",
        "https://docs.gitlab.com/ci/pipelines/compute_minutes/",
        "https://docs.gitlab.com/subscriptions/gitlab_com/compute_minutes/",
        "https://docs.gitlab.com/user/gitlab_com/",
        "https://about.gitlab.com/pricing/",
        "https://about.gitlab.com/company/visiting/"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "buildkite-hosted-agents",
      "name": "Buildkite hosted agents",
      "company": "Buildkite Pty Ltd",
      "hq": null,
      "kind": "platform",
      "category": "platform-hosted",
      "website": "https://buildkite.com",
      "docs": "https://buildkite.com/docs/agent/buildkite-hosted",
      "pricing_url": "https://buildkite.com/pricing/",
      "github": "https://github.com/buildkite/agent",
      "tagline": "Buildkite’s managed agents: a single-use virtualized environment per job on Linux, macOS and Windows, billed per vCPU minute.",
      "ci_systems": [
        "Buildkite"
      ],
      "os": [
        "Linux",
        "macOS",
        "Windows"
      ],
      "arch": [
        "x64",
        "arm64"
      ],
      "isolation": "Each job runs in its own virtualized environment that is destroyed with its data after the job; Buildkite describes hypervisor-level isolation between instances on hardware shared across customers",
      "isolation_short": "Ephemeral virtualized environment",
      "isolation_class": "vm",
      "ephemeral": true,
      "open_source": false,
      "license": null,
      "self_host": false,
      "byoc": false,
      "clouds": null,
      "gpu": null,
      "gpu_detail": null,
      "hardware": "Buildkite says it uses latest-generation Mac (M4) and AMD Zen-based hardware in a private cloud, with cache volumes on local NVMe-attached disks",
      "performance": "Buildkite says hosted agents deliver up to 3x faster performance than equivalent sized machines from other CI/CD providers and cloud platforms (vendor claim)",
      "sizes": "Linux AMD64 and ARM64: 2 vCPU / 4 GB to 64 vCPU / 256 GB (16 vCPU and above on Enterprise); macOS M4: 6 vCPU / 28 GB and 12 vCPU / 56 GB; Windows AMD64: 2 vCPU / 8 GB to 16 vCPU / 64 GB",
      "max_size_short": "64 vCPU / 256 GB",
      "cache": "Cache volumes and Git mirror volumes on Linux and macOS at no additional cost, retained for up to 14 days from last use, plus an internal container registry per cluster. Remote Docker builders with Docker layer caching are available on the Enterprise plan.",
      "docker_layer_cache": true,
      "concurrency": "Set by plan as combined vCPUs in use: on Pro up to 256 Linux AMD64, 64 Linux ARM64, 48 M4 Mac and 64 Windows vCPU; on Free up to 20 Linux AMD64 vCPU; custom on Enterprise",
      "job_limits": "Linux instances run for up to 8 hours and macOS instances for up to 4 hours; longer on request to Buildkite support",
      "static_ips": true,
      "ssh_access": true,
      "observability": "The Usage page of the Buildkite organization shows vCPU minutes, or job minutes by instance shape.",
      "setup": "Create a Buildkite hosted queue in a cluster, choosing the machine type and instance shape, then target that queue from your pipeline steps.",
      "runs_on": "queue: macos-medium",
      "pricing": "Pay-as-you-go on the Pro plan ($30 per active user per month): Linux $0.004 per vCPU minute on AMD64 and ARM64 ($0.008/min for 2 vCPU, $0.016 for 4, $0.032 for 8), Mac $0.02 per vCPU minute ($0.12/min for M4 Medium, $0.24 for M4 Large), Windows $0.008 per vCPU minute. Metered to the second with no minimums; cache storage is not charged.",
      "price_short": "$0.008/min",
      "free_tier": "Free plan: up to 2,000 Linux vCPU minutes per month on the Small AMD64 shape. Pro includes 4,000 Linux AMD64 vCPU minutes per month. A 30-day all-access trial needs no credit card",
      "compliance": "SOC 2 Type II",
      "regions": "US East Coast data centers only",
      "best_for": "Teams on Buildkite Pipelines that want Buildkite to operate the agents, including Apple silicon macOS machines",
      "overview": [
        "**Buildkite hosted agents** are agents that Buildkite runs for Buildkite Pipelines, as an alternative to operating your own. You create a hosted queue in a cluster, pick a machine type and instance shape, and jobs sent to that queue each start a new virtualized environment that is destroyed when the job ends. Linux jobs run from an agent image you can customize; macOS jobs run in VMs on dedicated Mac hardware; Windows agents are in public preview.",
        "Billing is per **vCPU minute**, metered to the second, with separate rates for Linux, Mac and Windows. Cache volumes on NVMe storage, Git mirroring and an internal container registry are included, and Enterprise customers get remote Docker builders. Linux and macOS jobs offer browser terminal access for debugging. The service runs only in US East Coast data centers, and macOS agents require a Pro or Enterprise plan."
      ],
      "notable": [
        "Buildkite says the company was founded in Australia in 2013",
        "The M2 macOS instance shapes were deprecated and removed on July 1, 2025, leaving M4 shapes",
        "As of November 2026 Windows hosted agents are documented as a public preview on Trial, Pro and Enterprise plans",
        "As of November 2026 Buildkite documents a public preview that runs GitHub Actions workflow files as Buildkite Pipelines jobs"
      ],
      "pros": [
        "You use Buildkite Pipelines and want agents that Buildkite provisions and scales",
        "You need Apple silicon M4 macOS agents with terminal and browser-based desktop access",
        "You want cache volumes and Git mirrors without a separate storage charge",
        "You want billing by vCPU minute, metered to the second with no minimum charge"
      ],
      "cons": [
        "Hosted exclusively in US East Coast data centers",
        "Signed pipeline validation is not supported for hosted agents",
        "Windows hosted agents are in public preview, without cache volumes, terminal access or custom base images",
        "macOS agents need a Pro or Enterprise plan, and ARM64 Linux shapes are not available on the Free plan"
      ],
      "verdict": "Buildkite’s own managed agents, suited to Buildkite Pipelines users who want hosted Linux, macOS and Windows capacity billed by vCPU minute.",
      "use_cases": {
        "macos-ios": "macOS agents run on M4 Mac hardware with selectable macOS and Xcode versions, and iOS build guidance is documented.",
        "arm64": "Linux ARM64 instance shapes from 2 to 64 vCPU are offered at the same per-vCPU rate as AMD64.",
        "docker-builds": "Container cache volumes, an internal container registry and Enterprise remote Docker builders with layer caching are documented.",
        "windows": "Windows Server 2022 hosted agents are available in public preview on AMD64."
      },
      "peers": [
        "buildkite-elastic-ci-stack",
        "github-hosted-runners",
        "gitlab-hosted-runners",
        "azure-pipelines-agents",
        "aws-codebuild-runners",
        "macstadium-orka"
      ],
      "sources": [
        "https://buildkite.com/docs/agent/buildkite-hosted",
        "https://buildkite.com/docs/agent/buildkite-hosted/linux",
        "https://buildkite.com/docs/agent/buildkite-hosted/macos",
        "https://buildkite.com/docs/agent/buildkite-hosted/windows",
        "https://buildkite.com/docs/agent/buildkite-hosted/cache-volumes",
        "https://buildkite.com/docs/agent/buildkite-hosted/terminal-access",
        "https://buildkite.com/docs/agent/buildkite-hosted/network-security",
        "https://buildkite.com/docs/pipelines/migration/run-github-actions-workflows",
        "https://buildkite.com/pricing/",
        "https://buildkite.com/about/company/",
        "https://buildkite.com/about/security/"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "azure-pipelines-agents",
      "name": "Microsoft-hosted agents",
      "company": "Microsoft Corporation",
      "hq": {
        "city": "Redmond",
        "region": "Washington",
        "country": "US",
        "source": "https://www.microsoft.com/en-us/privacy/privacystatement"
      },
      "kind": "platform",
      "category": "platform-hosted",
      "website": "https://azure.microsoft.com/en-us/products/devops/pipelines",
      "docs": "https://learn.microsoft.com/en-us/azure/devops/pipelines/agents/hosted",
      "pricing_url": "https://azure.microsoft.com/en-us/pricing/details/devops/azure-devops-services/",
      "github": "https://github.com/microsoft/azure-pipelines-agent",
      "tagline": "Azure Pipelines’ built-in agents: a fresh 2-core VM per job on Windows, Linux or macOS, sold per parallel job rather than per minute.",
      "ci_systems": [
        "Azure Pipelines"
      ],
      "os": [
        "Linux",
        "macOS",
        "Windows"
      ],
      "arch": [
        "x64"
      ],
      "isolation": "Each job gets a freshly imaged virtual machine that is discarded after one job; each agent is dedicated to a single organization and each VM hosts a single agent",
      "isolation_short": "Fresh VM",
      "isolation_class": "vm",
      "ephemeral": true,
      "open_source": false,
      "license": null,
      "self_host": false,
      "byoc": false,
      "clouds": null,
      "gpu": null,
      "gpu_detail": null,
      "hardware": "Windows and Linux agents run on Azure Standard_DS2_v2 general purpose VMs with 2 cores, 7 GB RAM and 14 GB SSD; macOS agents run on Mac Pros with a 3-core CPU and 14 GB RAM",
      "performance": null,
      "sizes": "One size: 2 vCPU / 7 GB on Windows and Linux and 3 cores / 14 GB on macOS. A separate pay-as-you-go GitHub-hosted agents pool offers 8- and 16-core Linux and Windows machines and Apple silicon macOS machines",
      "max_size_short": "2 vCPU / 7 GB",
      "cache": null,
      "docker_layer_cache": null,
      "concurrency": "One job at a time per Microsoft-hosted parallel job; the free tier provides one parallel job",
      "job_limits": "Free tier: 60 minutes per job and 1,800 minutes per month. Paid parallel jobs: up to 360 minutes (6 hours) per job. At least 10 GB of free disk space",
      "static_ips": false,
      "ssh_access": false,
      "observability": null,
      "setup": "Nothing to install: YAML pipelines default to the Azure Pipelines pool. Set pool vmImage to an image label; the free grant requires linking the organization to an Azure subscription.",
      "runs_on": "vmImage: ubuntu-latest",
      "pricing": "Sold per parallel job, not per minute: $40 per month for each extra Microsoft-hosted parallel job with unlimited minutes. The separate GitHub-hosted agents pool is billed per minute with no free minutes; Microsoft’s retail price list shows Linux 8-core $0.022, Linux 16-core $0.042, Windows 8-core $0.042, Windows 16-core $0.082, macOS Standard $0.062 and macOS XL $0.102 per minute.",
      "price_short": "$40/month per parallel job",
      "free_tier": "One free Microsoft-hosted parallel job with 1,800 minutes per month for private projects, limited to 60 minutes per job",
      "compliance": null,
      "regions": "Windows and Linux agents run in the Azure geography of the Azure DevOps organization (Asia Pacific, Australia, Canada, Europe, India, United Kingdom or United States); macOS agents always run in the US",
      "best_for": "Teams on Azure DevOps Services that want pipelines to run on standard images without managing agents",
      "overview": [
        "**Microsoft-hosted agents** are the agents Microsoft runs for Azure Pipelines in Azure DevOps Services, through the predefined pool named Azure Pipelines. A pipeline names a VM image such as `ubuntu-latest`, `windows-latest` or `macOS-latest`, and each job receives a freshly imaged virtual machine that is discarded afterwards. Microsoft says the VM images are updated regularly, about every three weeks.",
        "Capacity is bought as **parallel jobs** rather than minutes: one free job with 1,800 minutes a month, then $40 per month for each additional job with unlimited minutes. The machines are a fixed size that Microsoft says cannot be increased. For more powerful machines Microsoft points to a separate pay-as-you-go **GitHub-hosted agents** pool, to self-hosted agents, or to [Managed DevOps Pools](https://learn.microsoft.com/en-us/azure/devops/managed-devops-pools/overview)."
      ],
      "notable": [
        "The Ubuntu 26.04 and Windows Server 2025 with Visual Studio 2026 images became generally available on September 29, 2026",
        "Microsoft states that public projects are retired and that existing public projects convert to private in 2027",
        "The Windows Server 2019 image was retired on December 31, 2025",
        "As of November 2026 the macOS 15 ARM64 limited public preview on Microsoft-hosted agents is paused"
      ],
      "pros": [
        "You use Azure DevOps Services and want agents with no installation or maintenance",
        "You prefer a flat monthly price per parallel job with unlimited minutes",
        "You want each job on a freshly imaged VM dedicated to your organization",
        "You want Windows and Linux agents to run in the same Azure geography as your organization"
      ],
      "cons": [
        "Disk space and processing power cannot be increased on Microsoft-hosted agents",
        "No ability to connect remotely to an agent",
        "Private connections such as ExpressRoute or VPN cannot be used; traffic goes over the public network",
        "Hosted images do not conform to CIS hardening benchmarks"
      ],
      "verdict": "The built-in agents for Azure Pipelines, suited to Azure DevOps teams whose jobs fit a fixed 2-core machine and who prefer per-parallel-job pricing.",
      "use_cases": {
        "windows": "Windows Server 2022 and 2025 images with Visual Studio 2022 or 2026 are provided.",
        "macos-ios": "macOS 14, 15 and 26 images are provided on Mac Pro hardware that always runs in the US.",
        "security": "Microsoft documents that each VM is reimaged after each run and says hosted agents are safer than self-hosted agents for untrusted code such as fork contributions."
      },
      "peers": [
        "azure-managed-devops-pools",
        "github-hosted-runners",
        "gitlab-hosted-runners",
        "buildkite-hosted-agents",
        "aws-codebuild-runners"
      ],
      "sources": [
        "https://learn.microsoft.com/en-us/azure/devops/pipelines/agents/hosted?view=azure-devops",
        "https://learn.microsoft.com/en-us/azure/devops/pipelines/agents/github-hosted?view=azure-devops",
        "https://learn.microsoft.com/en-us/azure/devops/pipelines/licensing/concurrent-jobs?view=azure-devops",
        "https://azure.microsoft.com/en-us/pricing/details/devops/azure-devops-services/",
        "https://prices.azure.com/api/retail/prices?$filter=serviceName%20eq%20%27Azure%20DevOps%27",
        "https://github.com/microsoft/azure-pipelines-agent",
        "https://www.microsoft.com/en-us/privacy/privacystatement"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "aws-codebuild-runners",
      "name": "AWS CodeBuild-hosted runners",
      "company": "Amazon Web Services, Inc.",
      "hq": {
        "city": "Seattle",
        "region": "Washington",
        "country": "US",
        "source": "https://aws.amazon.com/privacy/"
      },
      "kind": "cloud",
      "category": "platform-hosted",
      "website": "https://aws.amazon.com/codebuild/",
      "docs": "https://docs.aws.amazon.com/codebuild/latest/userguide/runners.html",
      "pricing_url": "https://aws.amazon.com/codebuild/pricing/",
      "github": null,
      "tagline": "AWS CodeBuild projects that start a single-use runner in your AWS account for each GitHub Actions, GitLab or Buildkite job.",
      "ci_systems": [
        "GitHub Actions",
        "GitLab CI",
        "Buildkite"
      ],
      "os": [
        "Linux",
        "Windows"
      ],
      "arch": [
        "x64",
        "arm64"
      ],
      "isolation": "Each job starts a CodeBuild build in a fresh environment that is discarded when it finishes; runner builds use CodeBuild containers or instances on Amazon EC2, or AWS Lambda compute",
      "isolation_short": "Fresh build environment",
      "isolation_class": "mixed",
      "ephemeral": true,
      "open_source": false,
      "license": null,
      "self_host": false,
      "byoc": true,
      "clouds": [
        "AWS"
      ],
      "gpu": true,
      "gpu_detail": "Linux GPU compute types gpu_small (4 vCPU, 16 GiB) at $0.05 and gpu_large (32 vCPU, 255 GiB) at $0.30 per build minute in US East (N. Virginia); GPU models are not stated on the pages checked.",
      "hardware": "On-demand Amazon EC2 or AWS Lambda compute, or reserved capacity fleets, on x86-64 and Arm; AWS says runners can use its latest instance types, including ARM-based instances",
      "performance": null,
      "sizes": "On-demand EC2: 2 vCPU / 4 GiB to 72 vCPU / 144 GiB on Linux x86-64, 2 vCPU / 4 GiB to 48 vCPU / 96 GiB on Arm, 4 vCPU / 8 GiB to 72 vCPU / 144 GiB on Windows. Lambda: 1 to 10 GiB of memory",
      "max_size_short": "72 vCPU / 144 GB",
      "cache": "CodeBuild projects can cache to Amazon S3 or use a local cache with source, Docker layer and custom modes. A Docker image server with a shared layer cache is billed separately per second.",
      "docker_layer_cache": true,
      "concurrency": "Concurrently running builds are capped by per-Region service quotas for each environment type, which AWS lists as adjustable",
      "job_limits": "Maximum build timeout of 2,160 minutes (36 hours)",
      "static_ips": null,
      "ssh_access": null,
      "observability": "AWS CloudTrail integration is documented for runner projects.",
      "setup": "Create a CodeBuild project with a webhook for WORKFLOW_JOB_QUEUED events (or the GitLab or Buildkite equivalent), then reference the project name in runs-on, tags or agents. Image, instance size and fleet can be overridden with labels.",
      "runs_on": "codebuild-myProject-${{ github.run_id }}-${{ github.run_attempt }}",
      "pricing": "Standard CodeBuild rates per build minute in US East (N. Virginia): Linux x86-64 small (2 vCPU) $0.005, medium $0.01, large $0.02; Arm small $0.0034, medium $0.007, large $0.015; Windows medium $0.018, large $0.036. EC2 builds are rounded up to the minute; Lambda compute is billed per second. Rates vary by Region.",
      "price_short": "$0.005/min",
      "free_tier": "100 build minutes per month on general1.small or arm1.small on-demand EC2, or 6,000 build seconds per month on 1 GB Lambda compute",
      "compliance": null,
      "regions": "All AWS Regions where CodeBuild is available",
      "best_for": "Teams on AWS that want GitHub Actions, GitLab or Buildkite jobs to run inside their own account with IAM and VPC access",
      "overview": [
        "**AWS CodeBuild-hosted runners** let a CodeBuild project act as a self-hosted runner for GitHub Actions, GitLab CI/CD or Buildkite. A webhook on the project receives a job-queued event, and CodeBuild starts a build in your AWS account that registers a single-use runner, runs the job and is then discarded. Workflows target the project by name in `runs-on` (or GitLab `tags`, or Buildkite `agents`), with optional labels to override the image, instance size or fleet.",
        "Because the runner is an ordinary CodeBuild build, jobs get what AWS describes as native integration with IAM, AWS Secrets Manager, AWS CloudTrail and Amazon VPC, and are billed at standard CodeBuild rates with no separate runner fee. Compute can be on-demand EC2, AWS Lambda or a reserved capacity fleet. Documented runner platforms are Amazon Linux, Ubuntu and Windows Server Core on x86-64 and Arm."
      ],
      "notable": [
        "AWS announced managed GitHub Actions self-hosted runners in CodeBuild on April 24, 2024",
        "AWS announced managed GitLab self-hosted runners in CodeBuild on September 17, 2024",
        "AWS announced native support for self-hosted Buildkite runners in CodeBuild on February 3, 2025"
      ],
      "pros": [
        "You want CI jobs to run inside your own AWS account with IAM roles and VPC access",
        "You use GitHub Actions, GitLab or Buildkite and want one runner service for all of them",
        "You want single-use runners without operating an autoscaler",
        "You need Arm, GPU or Windows compute billed at standard CodeBuild rates"
      ],
      "cons": [
        "macOS is not listed among the supported runner platforms, and on-demand fleets do not support macOS",
        "Build duration on EC2 compute is rounded up to the nearest minute",
        "Buildkite webhooks must be created manually; CloudFormation is not supported for them",
        "Secrets printed in the log are not masked by default"
      ],
      "verdict": "AWS’s managed way to run GitHub Actions, GitLab and Buildkite jobs on single-use CodeBuild compute in your own account, suited to teams already on AWS.",
      "use_cases": {
        "byoc": "Runner builds execute in your AWS account, with access to IAM, Secrets Manager, CloudTrail and Amazon VPC.",
        "arm64": "Arm compute types and Amazon Linux aarch64 images are supported for runner builds.",
        "gitlab": "CodeBuild can host self-managed GitLab runners triggered by a GitLab webhook.",
        "gpu": "Linux GPU compute types can be selected with the instance-size label.",
        "security": "Each job runs in a fresh build environment that AWS says is isolated from other users and discarded on completion."
      },
      "peers": [
        "runs-on",
        "terraform-aws-github-runner",
        "github-hosted-runners",
        "gitlab-hosted-runners",
        "buildkite-elastic-ci-stack",
        "azure-managed-devops-pools",
        "actions-runner-controller"
      ],
      "sources": [
        "https://docs.aws.amazon.com/codebuild/latest/userguide/runners.html",
        "https://docs.aws.amazon.com/codebuild/latest/userguide/action-runner-overview.html",
        "https://docs.aws.amazon.com/codebuild/latest/userguide/action-runner-questions.html",
        "https://docs.aws.amazon.com/codebuild/latest/userguide/sample-github-action-runners-update-labels.html",
        "https://docs.aws.amazon.com/codebuild/latest/userguide/sample-github-action-runners-update-yaml.images.html",
        "https://docs.aws.amazon.com/codebuild/latest/userguide/gitlab-runner.html",
        "https://docs.aws.amazon.com/codebuild/latest/userguide/buildkite-runner.html",
        "https://docs.aws.amazon.com/codebuild/latest/userguide/buildkite-runner-about.html",
        "https://docs.aws.amazon.com/codebuild/latest/userguide/build-env-ref-compute-types.html",
        "https://docs.aws.amazon.com/codebuild/latest/userguide/fleets.html",
        "https://docs.aws.amazon.com/codebuild/latest/userguide/build-caching.html",
        "https://docs.aws.amazon.com/codebuild/latest/userguide/limits.html",
        "https://aws.amazon.com/codebuild/pricing/",
        "https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/CodeBuild/current/us-east-1/index.json",
        "https://aws.amazon.com/codebuild/faqs/",
        "https://aws.amazon.com/about-aws/whats-new/2024/04/aws-codebuild-managed-github-action-runners/",
        "https://aws.amazon.com/about-aws/whats-new/2024/09/aws-codebuild-managed-gitlab-runners/",
        "https://aws.amazon.com/about-aws/whats-new/2025/02/aws-codebuild-integrates-buildkite/",
        "https://aws.amazon.com/privacy/"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "azure-managed-devops-pools",
      "name": "Managed DevOps Pools",
      "company": "Microsoft Corporation",
      "hq": {
        "city": "Redmond",
        "region": "Washington",
        "country": "US",
        "source": "https://www.microsoft.com/en-us/privacy/privacystatement"
      },
      "kind": "platform",
      "category": "platform-hosted",
      "website": "https://learn.microsoft.com/en-us/azure/devops/managed-devops-pools/overview",
      "docs": "https://learn.microsoft.com/en-us/azure/devops/managed-devops-pools/overview",
      "pricing_url": "https://learn.microsoft.com/en-us/azure/devops/managed-devops-pools/pricing",
      "github": null,
      "tagline": "Azure service that runs autoscaled Azure Pipelines agents on VM sizes and images you choose, billed at standard Azure rates.",
      "ci_systems": [
        "Azure Pipelines"
      ],
      "os": [
        "Linux",
        "Windows"
      ],
      "arch": [
        "x64"
      ],
      "isolation": "Agents are Azure VMs in a Microsoft-owned subscription, with each pool isolated in a dedicated virtual network; stateless pools give every job a new agent, stateful pools reuse agents for up to seven days",
      "isolation_short": "Azure VM",
      "isolation_class": "vm",
      "ephemeral": null,
      "open_source": false,
      "license": null,
      "self_host": false,
      "byoc": false,
      "clouds": [
        "Azure"
      ],
      "gpu": null,
      "gpu_detail": null,
      "hardware": "Azure virtual machine sizes that you select for the pool, up to five sizes per pool with Instance Mix, with an optional attached data disk",
      "performance": null,
      "sizes": "Azure VM sizes of your choice, subject to Managed DevOps Pools quota for the VM family",
      "max_size_short": null,
      "cache": "Stateful pools keep agent state for up to seven days so later jobs can reuse packages and files from earlier runs. Standby agents can be kept online on a manual or automatic schedule.",
      "docker_layer_cache": null,
      "concurrency": "Limited by the pool’s Maximum agents setting, the Managed DevOps Pools quota for the VM family, and the number of self-hosted parallel jobs in the Azure DevOps organization",
      "job_limits": "Individual jobs can run for a maximum of two days",
      "static_ips": true,
      "ssh_access": false,
      "observability": "Predefined metrics charts, an Agents pane showing provisioning status, and optional log export to Azure Log Analytics.",
      "setup": "Register the Microsoft.DevOpsInfrastructure resource provider, create a pool with a Dev Center project in the Azure portal, ARM, Bicep or Azure CLI, then reference the pool by name in your pipelines.",
      "runs_on": "pool: name: fabrikam-managed-pool",
      "pricing": "No per-minute price. You pay standard Azure rates for the compute, storage and data egress the pool uses, plus Azure DevOps self-hosted parallel jobs: one is free and each extra one costs $15 per month. Azure Hybrid Benefit cannot be used.",
      "price_short": "$15/month per parallel job + Azure VM cost",
      "free_tier": "One free self-hosted parallel job with unlimited minutes per Azure DevOps organization; Azure compute is still charged",
      "compliance": null,
      "regions": "Azure public cloud regions listed for the Microsoft.DevOpsInfrastructure resource provider in your subscription",
      "best_for": "Azure DevOps teams that need bigger or custom-image agents and private network access without operating scale sets",
      "overview": [
        "**Managed DevOps Pools** is an Azure service that creates and scales agent pools for Azure Pipelines. Microsoft describes it as an evolution of Virtual Machine Scale Sets agent pools. You create a pool as an Azure resource, choose VM sizes, images (Azure Pipelines images, marketplace images or your own from Azure Compute Gallery) and a maximum agent count, and the service provisions agents on demand or keeps standby agents ready on a schedule.",
        "The agent VMs run in a subscription owned by Microsoft rather than your own, while charges for compute, storage and egress are billed at standard Azure rates. A pool can use a Microsoft-provided network or a subnet you delegate from your own virtual network to reach private resources. Pools are **stateless** by default, giving each job a new agent, or **stateful**, reusing agents for up to seven days."
      ],
      "notable": [
        "Microsoft’s features timeline lists the general-availability feature set, including more Azure regions and Key Vault integration, under November 2024",
        "Support for public static IP addresses was added in November 2025",
        "Generation 2 versions of the Azure Pipelines images were added in April 2026",
        "Instance Mix, allowing up to five VM sizes per pool, was added in September 2026"
      ],
      "pros": [
        "You need Azure Pipelines agents more powerful than Microsoft-hosted agents",
        "You want agents attached to your own virtual network to reach private resources",
        "You want to use your own VM images or the same images as Microsoft-hosted agents",
        "You need jobs that run longer than six hours, up to two days",
        "You want agents that keep state between jobs for up to seven days"
      ],
      "cons": [
        "Agent VMs run in a Microsoft-owned subscription and cannot currently be accessed directly",
        "Azure Hybrid Benefit cannot be used with Managed DevOps Pools",
        "Without standby agents, a pipeline might wait up to 15 minutes for an agent to be provisioned",
        "Supported only with Azure DevOps Services in the Azure public cloud"
      ],
      "verdict": "Microsoft’s managed alternative to scale-set agents for Azure Pipelines, suited to teams that need custom sizes, images and private networking at Azure compute rates.",
      "use_cases": {
        "byoc": "Pools are Azure resources billed through your subscription and can attach agents to a subnet of your own virtual network, although the VMs run in a Microsoft-owned subscription.",
        "windows": "Windows Server 2022 and 2025 Azure Pipelines images are provided.",
        "security": "Stateless pools give every job a new agent, and Microsoft recommends them as a defense against supply chain attacks.",
        "faster-builds": "You can select larger Azure VM sizes and keep agent state for up to seven days for cache hits."
      },
      "peers": [
        "azure-pipelines-agents",
        "aws-codebuild-runners",
        "github-hosted-runners",
        "gitlab-hosted-runners",
        "buildkite-elastic-ci-stack"
      ],
      "sources": [
        "https://learn.microsoft.com/en-us/azure/devops/managed-devops-pools/overview?view=azure-devops",
        "https://learn.microsoft.com/en-us/azure/devops/managed-devops-pools/architecture-overview?view=azure-devops",
        "https://learn.microsoft.com/en-us/azure/devops/managed-devops-pools/pricing?view=azure-devops",
        "https://learn.microsoft.com/en-us/azure/devops/managed-devops-pools/features-timeline?view=azure-devops",
        "https://learn.microsoft.com/en-us/azure/devops/managed-devops-pools/configure-scaling?view=azure-devops",
        "https://learn.microsoft.com/en-us/azure/devops/managed-devops-pools/configure-images?view=azure-devops",
        "https://learn.microsoft.com/en-us/azure/devops/managed-devops-pools/configure-pool-settings?view=azure-devops",
        "https://learn.microsoft.com/en-us/azure/devops/managed-devops-pools/configure-networking?view=azure-devops",
        "https://azure.microsoft.com/en-us/pricing/details/devops/azure-devops-services/",
        "https://www.microsoft.com/en-us/privacy/privacystatement"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "runs-on",
      "name": "RunsOn",
      "company": "RunsOn (Cyril Rohr EIRL)",
      "hq": {
        "city": "Thorigné-Fouillard",
        "region": "Brittany",
        "country": "FR",
        "source": "https://runs-on.com/legal/privacy/"
      },
      "kind": "byoc",
      "category": "your-cloud",
      "website": "https://runs-on.com",
      "docs": "https://runs-on.com/docs/",
      "pricing_url": "https://runs-on.com/pricing/",
      "github": "https://github.com/runs-on/runs-on",
      "tagline": "Self-hosted GitHub Actions runner control plane installed in your own AWS account, launching one fresh EC2 instance per job.",
      "ci_systems": [
        "GitHub Actions"
      ],
      "os": [
        "Linux",
        "Windows"
      ],
      "arch": [
        "x64",
        "arm64"
      ],
      "isolation": "Each job runs on an EC2 instance that no other job has used and that is terminated when the job ends",
      "isolation_short": "EC2 instance",
      "isolation_class": "vm",
      "ephemeral": true,
      "open_source": false,
      "license": null,
      "self_host": true,
      "byoc": true,
      "clouds": [
        "AWS"
      ],
      "gpu": true,
      "gpu_detail": "GPU images with NVIDIA drivers are published for x64 and Graviton instances (examples use g4dn, g5 and g5g); GPU compute is billed by AWS and no RunsOn GPU price is published.",
      "hardware": "Any EC2 instance family selected by label, including Graviton and families with local NVMe; the docs cite 1 to 896 vCPUs.",
      "performance": "RunsOn says Rust builds run 2.0x, TypeScript 2.1x and Docker builds 2.6x faster than on GitHub-hosted runners in its own benchmarks (vendor claim)",
      "sizes": "Predefined runners from 1 to 64 vCPU on x64 and arm64; labels can request any EC2 instance type, which the docs describe as 1 to 896 vCPUs",
      "max_size_short": "896 vCPU",
      "cache": "Magic Cache redirects actions/cache and Buildx type=gha to an S3 bucket in your account with no size cap (extras=s3-cache). Docker layers can also be cached in a per-stack ECR registry or on sticky EBS disks, and v3.6.0 added S3-backed remote caches for Bazel, Go, Gradle, sccache and Turborepo.",
      "docker_layer_cache": true,
      "concurrency": "No RunsOn concurrency limit is documented; the Flex mode is recommended for up to about 50,000 jobs a day",
      "job_limits": "Flex terminates runners after 12 hours by default (RunnerMaxRuntime, configurable, minimum 30 minutes)",
      "static_ips": true,
      "ssh_access": true,
      "observability": "Per-job cost estimates and runner metrics, an automatic CloudWatch dashboard, OpenTelemetry export of logs, metrics and traces, and a daily budget alarm; email cost reports on Flex.",
      "setup": "Deploy the CloudFormation stack (or the Terraform/OpenTofu module) in your AWS account, register the private GitHub App from the stack output, then change runs-on in each workflow.",
      "runs_on": "runs-on=${{ github.run_id }}/runner=2cpu-linux-x64",
      "pricing": "Flat annual licence per legal entity: Starter €300/year (under 50k jobs/month), Growth €900/year (under 200k), Scale €1,800/year (under 500k), Sponsorship €3,600/year (unlimited jobs, source access). EC2, EBS and S3 are billed by AWS with no RunsOn markup; RunsOn says the control plane costs from about $7/month.",
      "price_short": "€300/yr licence + AWS cost",
      "free_tier": "Free licence for non-commercial use (open source, non-profit, educational, personal) with a public link back; 15-day trial on Starter, Growth and Scale",
      "compliance": null,
      "regions": null,
      "best_for": "Teams on GitHub Actions that want ephemeral EC2 runners, caches and networking kept inside their own AWS account under a flat annual licence",
      "overview": [
        "**RunsOn** is a runner control plane for GitHub Actions that you install in your own AWS account with a CloudFormation stack or a Terraform/OpenTofu module. The control plane runs on Fargate, and every job gets a new EC2 instance that is terminated when the job ends. Linux (x64 and arm64) and Windows images are provided; macOS jobs stay on GitHub-hosted runners. Spot instances are used by default, with on-demand fallback.",
        "Two modes share one licence. **Flex** lets each workflow choose CPU, RAM, instance family, image and disk in its runs-on label; **Fleet**, added in June 2026, uses a Terraform-managed catalogue built on GitHub runner scale sets. The CloudFormation templates are MIT-licensed, while the server and agent are proprietary with source access on the top tier. Pricing is a [flat annual licence](https://runs-on.com/pricing/) by job volume, and compute is billed by AWS."
      ],
      "notable": [
        "RunsOn launched in 2023, according to its about page, which describes it as funded by customer licences rather than investors.",
        "v3.1.0 (2 June 2026) introduced Fleet, a pool-based mode built on GitHub Actions runner scale sets.",
        "v3.6.0 (4 November 2026) added remote caches for Bazel, Go, Gradle, sccache and Turborepo served from your own S3 bucket.",
        "GitHub Enterprise Server support arrived in v2.7.2 (3 April 2025)."
      ],
      "pros": [
        "You want every job on a fresh EC2 instance inside your own AWS account, with code, secrets, caches and logs staying there",
        "You prefer a flat annual licence by job volume over per-minute fees, with compute billed directly by AWS",
        "You need Linux x64, arm64, Windows or GPU runners chosen per job through labels",
        "You need static egress IPs, private subnets, or SSM and SSH access to running jobs",
        "You want S3-backed Actions, Docker layer and build-tool caches without a size cap"
      ],
      "cons": [
        "AWS only, and macOS is not supported: macOS jobs stay on GitHub-hosted runners",
        "The server and agent are proprietary; full source is offered only with the €3,600/year Sponsorship licence",
        "You run the stack yourself and pay AWS for the control plane, any NAT gateway and all compute",
        "A licence check and one event per job are sent to runs-on.com, and the job event cannot be turned off"
      ],
      "verdict": "RunsOn is licensed software that runs ephemeral GitHub Actions runners entirely in your own AWS account, suited to teams that want AWS-native control and a flat annual fee.",
      "use_cases": {
        "byoc": "The control plane and every runner run in your own AWS account, installed by CloudFormation or Terraform.",
        "lower-cost": "RunsOn charges a flat annual licence and compute is billed by AWS at spot or on-demand rates with no markup.",
        "docker-builds": "Docker layers can be cached through S3-backed Buildx caching, a per-stack ECR registry or sticky EBS disks.",
        "arm64": "Ubuntu arm64 images run on Graviton instance families selected by label.",
        "windows": "Windows Server 2025 images are provided for x64 runners.",
        "gpu": "GPU images with NVIDIA drivers and CUDA are published for x64 and Graviton GPU instances.",
        "security": "Runners can be confined to private subnets with static egress IPs, and each job gets a single-use instance."
      },
      "peers": [
        "sprinters",
        "hyperenv",
        "cirun",
        "terraform-aws-github-runner",
        "aws-codebuild-runners",
        "actions-runner-controller",
        "warpbuild",
        "actuated"
      ],
      "sources": [
        "https://runs-on.com/",
        "https://runs-on.com/pricing/",
        "https://runs-on.com/docs/",
        "https://runs-on.com/docs/flex-vs-fleet/",
        "https://runs-on.com/security/",
        "https://runs-on.com/docs/runners/platforms/linux/",
        "https://runs-on.com/docs/runners/platforms/windows/",
        "https://runs-on.com/runners/gpu/",
        "https://runs-on.com/docs/runners/labels/",
        "https://runs-on.com/docs/runners/capabilities/static-ips/",
        "https://runs-on.com/docs/runners/capabilities/remote-access/",
        "https://runs-on.com/docs/performance/caching/docker/",
        "https://runs-on.com/docs/observability/",
        "https://runs-on.com/docs/control-plane/housekeeping/",
        "https://runs-on.com/changelog/",
        "https://runs-on.com/about/",
        "https://runs-on.com/legal/privacy/",
        "https://github.com/runs-on/runs-on",
        "https://github.com/runs-on/runs-on/blob/main/SPONSORSHIP.md"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "cirun",
      "name": "Cirun",
      "company": "Aktech Labs",
      "hq": {
        "city": "London",
        "region": "England",
        "country": "GB",
        "source": "https://aktechlabs.com/about/"
      },
      "kind": "byoc",
      "category": "your-cloud",
      "website": "https://cirun.io",
      "docs": "https://docs.cirun.io",
      "pricing_url": "https://cirun.io/pricing",
      "github": "https://github.com/cirunlabs/cirun-agent",
      "tagline": "Hosted control plane that launches on-demand GitHub Actions runners in your own cloud accounts or on-prem hardware from a .cirun.yml file.",
      "ci_systems": [
        "GitHub Actions"
      ],
      "os": [
        "Linux",
        "macOS",
        "Windows"
      ],
      "arch": [
        "x64",
        "arm64"
      ],
      "isolation": "Each job runs in a fresh virtual machine created in your cloud account and destroyed after the job; on-prem runners use cloud-hypervisor VMs on Linux, Apple Virtualization VMs on macOS or opt-in Docker containers",
      "isolation_short": "Cloud VM",
      "isolation_class": "vm",
      "ephemeral": true,
      "open_source": false,
      "license": null,
      "self_host": null,
      "byoc": true,
      "clouds": [
        "AWS",
        "Google Cloud",
        "Azure",
        "Hetzner",
        "DigitalOcean",
        "Oracle Cloud",
        "OpenStack",
        "Vast.ai",
        "Bare metal"
      ],
      "gpu": true,
      "gpu_detail": "GPU runners are included in every plan; the docs show an NVIDIA Tesla T4 attached to a Google Cloud instance and list Vast.ai as a provider, with GPU cost billed by your cloud.",
      "hardware": "Any instance type offered by the connected cloud account, or your own Linux (KVM) and Apple Silicon machines through cirun-agent.",
      "performance": null,
      "sizes": null,
      "max_size_short": null,
      "cache": "Setting extra_config.cache: true redirects actions/cache to an S3 bucket in your own AWS account with no Cirun size limit (automatic mode covers Linux runners on AWS); a separate cirunlabs/cache/s3 action works with any S3-compatible store. Docker Buildx type=gha caching through the same proxy is documented.",
      "docker_layer_cache": true,
      "concurrency": "No plan concurrency limit is published; the on-prem agent defaults to 2 concurrent runners on macOS hosts and is unlimited by default on Linux",
      "job_limits": null,
      "static_ips": null,
      "ssh_access": null,
      "observability": null,
      "setup": "Install the Cirun GitHub App, add cloud credentials in the dashboard, commit a .cirun.yml that defines runners, then point runs-on at the runner label.",
      "runs_on": "cirun-gpu-runner--${{ github.run_id }}",
      "pricing": "Open Source: $0 for unlimited public repositories. Startup: $29 per month for 3 private repositories. Business: $79 per month for 10 private repositories. Enterprise: custom pricing. Compute is billed by your own cloud provider.",
      "price_short": "$29/mo (3 private repos) + cloud cost",
      "free_tier": "Free for public repositories, with unlimited runners, ARM and GPU runners on all supported cloud providers",
      "compliance": null,
      "regions": "Any region of the connected cloud account, set per runner in .cirun.yml",
      "best_for": "Teams and open-source projects on GitHub Actions that want on-demand runners, including GPU and ARM machines, across several clouds or on their own hardware",
      "overview": [
        "**Cirun** is a hosted service that creates GitHub Actions runners in cloud accounts you own. You install its GitHub App, store cloud credentials in the dashboard and describe runners in a `.cirun.yml` file: cloud, instance type, machine image, region and labels. When a job with a matching label is queued, Cirun starts a virtual machine in your account and, according to its security page, destroys it immediately after the job completes.",
        "The docs name AWS, GCP, Azure, Hetzner, DigitalOcean, Oracle, OpenStack and Vast.ai as providers, and an MIT-licensed `cirun-agent` registers your own Linux or Apple Silicon machines as an on-prem runner pool. Options include spot or preemptible instances, custom images, Windows runners on AWS and Azure, scheduled idle runners and an S3-backed Actions cache. Pricing is a [flat monthly fee](https://cirun.io/pricing) by number of private repositories, and public repositories are free."
      ],
      "notable": [
        "The site footer carries a 2021-2026 copyright for Cirun.io, Aktech Labs.",
        "The homepage describes the company as profitable and bootstrapped (checked 2026-11-22).",
        "The caching docs list other clouds and Windows/macOS for automatic S3 caching as roadmap items (checked 2026-11-22)."
      ],
      "pros": [
        "You want runners in your own account on AWS, Google Cloud, Azure, DigitalOcean, Oracle Cloud, OpenStack or other listed providers from one configuration file",
        "You need GPU or ARM runners, which every plan includes",
        "You maintain public repositories, where the Open Source plan is free",
        "You want to register your own Linux or Apple Silicon machines as an on-prem runner pool",
        "You prefer a flat monthly price by private repository count instead of per-minute billing"
      ],
      "cons": [
        "Automatic S3 caching currently covers Linux runners on AWS only",
        "The AWS setup documents an IAM user access key with EC2 admin permissions entered in the Cirun dashboard",
        "Startup and Business plans are limited to 3 and 10 private repositories",
        "Windows runners are documented on AWS and Azure only and require a Windows licence with multi-tenant rights"
      ],
      "verdict": "Cirun is a hosted orchestrator that starts single-use GitHub Actions runners in your own cloud accounts or on your own machines, suited to teams that need several clouds, GPUs or ARM under a flat fee.",
      "use_cases": {
        "byoc": "Runners are created in cloud accounts you connect, or on your own hardware through cirun-agent.",
        "gpu": "GPU runners are included in every plan and the docs show GPU instances on Google Cloud and Vast.ai as a provider.",
        "arm64": "ARM-based runners are included in every plan and the on-prem Docker runner image is multi-arch.",
        "windows": "Windows runners are documented on AWS and Azure, including ARM64 on Azure.",
        "macos-ios": "The on-prem agent can run macOS VMs on Apple Silicon hosts through the lume executor."
      },
      "peers": [
        "runs-on",
        "actuated",
        "sprinters",
        "hyperenv",
        "warpbuild",
        "garm",
        "actions-runner-controller",
        "terraform-aws-github-runner"
      ],
      "sources": [
        "https://cirun.io/",
        "https://cirun.io/pricing",
        "https://docs.cirun.io/",
        "https://docs.cirun.io/quickstart",
        "https://docs.cirun.io/reference/yaml",
        "https://docs.cirun.io/cloud/",
        "https://docs.cirun.io/cloud/aws",
        "https://docs.cirun.io/on-prem",
        "https://docs.cirun.io/caching/",
        "https://docs.cirun.io/os-platform/windows",
        "https://docs.cirun.io/reference/idle-runners",
        "https://docs.cirun.io/security",
        "https://docs.cirun.io/terms-of-service",
        "https://github.com/cirunlabs/cirun-agent",
        "https://aktechlabs.com/about/"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "actuated",
      "name": "Actuated",
      "company": "OpenFaaS Ltd",
      "hq": {
        "city": "Peterborough",
        "region": "England",
        "country": "GB",
        "source": "https://github.com/self-actuated/actuated/blob/master/EULA.md"
      },
      "kind": "byoc",
      "category": "your-cloud",
      "website": "https://actuated.com",
      "docs": "https://docs.actuated.com",
      "pricing_url": "https://actuated.com/pricing",
      "github": "https://github.com/self-actuated/actuated",
      "tagline": "Managed control plane that runs each CI job in a Firecracker microVM on servers you provide, for GitHub Actions and GitLab CI.",
      "ci_systems": [
        "GitHub Actions",
        "GitLab CI"
      ],
      "os": [
        "Linux"
      ],
      "arch": [
        "x64",
        "arm64"
      ],
      "isolation": "Each job runs in its own Firecracker microVM on your server, created for one build and destroyed afterwards (Cloud Hypervisor is used on GPU hosts)",
      "isolation_short": "Firecracker microVM",
      "isolation_class": "microvm",
      "ephemeral": true,
      "open_source": false,
      "license": null,
      "self_host": false,
      "byoc": true,
      "clouds": [
        "Bare metal",
        "Hetzner",
        "AWS",
        "Google Cloud",
        "Azure",
        "DigitalOcean",
        "OpenStack",
        "VMware"
      ],
      "gpu": true,
      "gpu_detail": "GPUs on the host are passed through whole to one job at a time via VFIO, which requires switching the server from Firecracker to Cloud Hypervisor; jobs add a gpu label.",
      "hardware": "Your own bare-metal servers or VMs with nested virtualisation; the docs recommend NVMe storage and list Ampere, Graviton and Apple Silicon (Asahi Linux) hosts for Arm.",
      "performance": "Actuated says builds are often 2-3x faster, and cites a customer Arm build cut from about 6 hours under QEMU to 20 minutes (vendor claim)",
      "sizes": "No fixed sizes: vCPU and RAM are set per job in the label (for example actuated-4cpu-16gb) within the capacity of your server; fractional 0.25 and 0.5 vCPU sizes were announced in October 2025",
      "max_size_short": null,
      "cache": "A transparent cache announced in October 2025 redirects actions/cache, BuildKit and setup-* action caches to local NVMe on your server without workflow changes. The docs also cover a registry mirror and a self-hosted S3 cache (SeaweedFS or MinIO) on the host.",
      "docker_layer_cache": true,
      "concurrency": "The Self-Service plan lists unlimited concurrency; jobs queue once the RAM and CPU of your servers are allocated",
      "job_limits": null,
      "static_ips": null,
      "ssh_access": true,
      "observability": "Reports across organisation, repositories and users in the dashboard, Prometheus metrics for servers and jobs through a federation endpoint, and a CLI for agent and VM logs.",
      "setup": "Register your GitHub organisation, provision a server with KVM support, install the Actuated agent, then change runs-on to an actuated-* label.",
      "runs_on": "actuated-4cpu-16gb",
      "pricing": "Self-Service Plan: flat rate per server, $150 per month for the first and $125 for each additional, with unlimited build minutes, RAM/CPUs and concurrency. Servers are paid for separately. An October 2025 post says custom billing is available to customers needing more than 50 concurrent builds.",
      "price_short": "$150/mo per server + your servers",
      "free_tier": null,
      "compliance": null,
      "regions": "Wherever your servers are; the FAQ says the control plane is hosted in the United Kingdom",
      "best_for": "Teams that want microVM isolation for GitHub Actions or GitLab CI jobs on their own bare-metal or nested-virtualisation servers at a flat monthly rate",
      "overview": [
        "**Actuated** is a commercial service from OpenFaaS Ltd that turns servers you supply into CI runner hosts. You install its agent on a bare-metal machine or a VM with nested virtualisation, and its hosted control plane schedules a Firecracker microVM for each queued job. The VM boots a maintained Ubuntu image, runs one build with full root, Docker and Kubernetes available, and is then destroyed. Linux on x86_64 and 64-bit Arm is supported.",
        "Jobs choose their own vCPU and RAM through labels such as `actuated-4cpu-16gb` rather than fixed sizes. GitLab CI support was announced in July 2024 alongside GitHub Actions. The [pricing page](https://actuated.com/pricing) lists a flat monthly rate per server with unlimited minutes and concurrency, and the plan includes SSH debugging and Prometheus metrics. The docs list Hetzner, AWS bare metal, GCP, Azure, DigitalOcean and on-premises machines as host options."
      ],
      "notable": [
        "Launched in November 2022, according to the two-year review of 6 August 2024, which also announced the move to actuated.com.",
        "GitLab CI support was announced on 26 July 2024.",
        "A 31 October 2025 post announced auto enrollment, a transparent cache, fractional CPU sizes and custom billing.",
        "A 13 October 2025 post announced a Jenkins plugin for Firecracker builds, offered through SlicerVM.com subscriptions."
      ],
      "pros": [
        "You want each job isolated in its own Firecracker microVM without privileged containers or Docker-in-Docker",
        "You already have, or want to rent, bare-metal or nested-virtualisation servers and prefer a flat monthly rate per server",
        "You need native 64-bit Arm builds on hardware such as Ampere or Graviton hosts",
        "You run GitLab CI as well as GitHub Actions and want the same runner model for both",
        "You want to SSH into a running job to debug it"
      ],
      "cons": [
        "Windows and macOS are not supported",
        "Servers must be bare metal or support nested virtualisation, and Arm hosts must be bare metal",
        "You provision, pay for and maintain the servers, and jobs queue when they are full",
        "GitHub Enterprise Server is not supported out of the box"
      ],
      "verdict": "Actuated is a managed control plane with a host agent that runs each CI job in a single-use microVM on your own servers, suited to teams that want strong isolation and predictable flat pricing.",
      "use_cases": {
        "byoc": "Jobs run on servers you supply, from on-premises machines to Hetzner bare metal or cloud VMs with nested virtualisation.",
        "security": "Each job gets a single-use Firecracker microVM with no access to the host, and the vendor documents use with public repositories.",
        "arm64": "Arm jobs run natively on bare-metal Arm hosts such as Ampere Altra or Graviton without QEMU.",
        "gitlab": "GitLab CI is supported for GitLab.com and self-managed GitLab, with jobs opted in through an actuated tag.",
        "docker-builds": "Docker runs natively inside each microVM, and a transparent cache and registry mirror keep layers on local NVMe.",
        "gpu": "GPU passthrough to jobs is documented for hosts switched to Cloud Hypervisor."
      },
      "peers": [
        "fireactions",
        "cirun",
        "runs-on",
        "ubicloud",
        "actions-runner-controller",
        "gitlab-runner",
        "garm",
        "blacksmith"
      ],
      "sources": [
        "https://actuated.com/",
        "https://actuated.com/pricing",
        "https://actuated.com/gitlab",
        "https://actuated.com/blog",
        "https://actuated.com/blog/autoenrollment-caching-custom-billing-fractional-cpu",
        "https://actuated.com/blog/bringing-firecracker-to-jenkins",
        "https://docs.actuated.com/",
        "https://docs.actuated.com/faq/",
        "https://docs.actuated.com/provision-server/",
        "https://docs.actuated.com/test-build/",
        "https://docs.actuated.com/examples/gpu/",
        "https://docs.actuated.com/tasks/debug-ssh/",
        "https://docs.actuated.com/tasks/local-github-cache/",
        "https://docs.actuated.com/tasks/monitoring/",
        "https://docs.actuated.com/roadmap/",
        "https://github.com/self-actuated/actuated/blob/master/EULA.md"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "sprinters",
      "name": "Sprinters",
      "company": "InfrastructureX GmbH",
      "hq": {
        "city": "Munich",
        "region": "Bavaria",
        "country": "DE",
        "source": "https://sprinters.sh/impressum"
      },
      "kind": "byoc",
      "category": "your-cloud",
      "website": "https://sprinters.sh",
      "docs": "https://sprinters.sh/docs",
      "pricing_url": "https://sprinters.sh/pricing",
      "github": "https://github.com/sprinters-sh/sprinters",
      "tagline": "GitHub App that launches one ephemeral EC2 instance per GitHub Actions job in your own AWS account, charged per job rather than per minute.",
      "ci_systems": [
        "GitHub Actions"
      ],
      "os": [
        "Linux"
      ],
      "arch": [
        "x64",
        "arm64"
      ],
      "isolation": "Each job runs on a new EC2 instance in your VPC with a read-only boot volume, encrypted temp and swap volumes and a security group that prohibits ingress; the instance and its volumes are destroyed after the job",
      "isolation_short": "EC2 instance",
      "isolation_class": "vm",
      "ephemeral": true,
      "open_source": false,
      "license": null,
      "self_host": false,
      "byoc": true,
      "clouds": [
        "AWS"
      ],
      "gpu": null,
      "gpu_detail": null,
      "hardware": "612 EC2 instance types across Intel, AMD and Graviton families, with optional local NVMe or zram temp storage.",
      "performance": "Sprinters says jobs run 30% faster and about 10x cheaper per minute than on GitHub-hosted runners; its own Hugo benchmark shows 20% faster and 12.1x cheaper (vendor claim)",
      "sizes": "1 to 384 vCPUs, 0.5 to 3,072 GiB RAM and 1 to 65,536 GiB temp disk; defaults are t3.large on x64 and t4g.large on arm64",
      "max_size_short": "384 vCPU / 3072 GB",
      "cache": null,
      "docker_layer_cache": null,
      "concurrency": "The pricing page lists unlimited concurrency",
      "job_limits": "The pricing page lists unlimited build minutes per job",
      "static_ips": null,
      "ssh_access": null,
      "observability": "A Sprinters Console with job history and error details; temp disk utilisation is logged at the end of each job.",
      "setup": "Log in with GitHub and install the Sprinters GitHub App, launch the CloudFormation stack that creates a cross-account IAM role, then change runs-on in each workflow.",
      "runs_on": "sprinters:aws:ubuntu-latest",
      "pricing": "$0.01 per job on private repositories, counted only when the runner launches successfully; public repositories are free. There is no subscription or per-vCPU-minute fee; AWS bills compute, storage and bandwidth to your account.",
      "price_short": "$0.01/job + AWS cost",
      "free_tier": "Free for public repositories; 1,000 free jobs in the first 30 days for private repositories",
      "compliance": null,
      "regions": "10 AWS regions: us-east-1, us-west-2, ca-central-1, eu-central-1, eu-west-1, eu-west-2, me-central-1, ap-northeast-1, ap-northeast-2, ap-south-1",
      "best_for": "Teams on GitHub Actions that want Ubuntu runners on EC2 in their own AWS account with a per-job fee and no subscription",
      "overview": [
        "**Sprinters** is a GitHub App from InfrastructureX GmbH that runs GitHub Actions jobs on EC2 instances in your own AWS account. A CloudFormation stack creates a cross-account IAM role limited to seven actions; when a job with a `sprinters:aws` label is queued, the hosted platform launches one instance in your VPC, the runner registers with GitHub, and the instance and its volumes are destroyed when the job completes.",
        "Everything is set in the runs-on label: Ubuntu 22.04 or 24.04 images in full, slim and minimal variants, any of 612 instance types, region or subnet, spot mode and temp storage. The full images are described as identical to those on GitHub-hosted runners. Billing is [$0.01 per job](https://sprinters.sh/pricing) on private repositories with no per-minute fee, while AWS charges for the compute. Public repositories are free."
      ],
      "notable": [
        "Initial public release on 8 January 2025, per the changelog.",
        "Access to AWS resources in private subnets was added on 7 January 2026.",
        "Automatic on-demand rerun for spot jobs after instance reclaim was added on 27 March 2026.",
        "Support for the ap-northeast-2 (Seoul) region was added on 14 January 2026."
      ],
      "pros": [
        "You want a flat $0.01 per job fee with no subscription and no per-minute charge beyond your AWS bill",
        "You need runners inside your own VPC and subnets to reach private AWS resources",
        "You want to pick any of 612 x64 and arm64 EC2 instance types, up to 384 vCPUs, in the runs-on label",
        "You maintain public repositories, which are free",
        "You want images that match GitHub-hosted Ubuntu runners"
      ],
      "cons": [
        "Only Ubuntu 22.04 and 24.04 images are documented; Windows and macOS runners are not documented",
        "Runners can launch only in the 10 AWS regions the docs list",
        "The control plane runs on Sprinters’ infrastructure and uses a cross-account IAM role in your AWS account",
        "Caching features are not documented"
      ],
      "verdict": "Sprinters is a hosted orchestrator that launches single-use EC2 runners for GitHub Actions in your own AWS account, suited to teams that want per-job pricing and Ubuntu images matching GitHub’s.",
      "use_cases": {
        "byoc": "Runner instances launch in your own AWS account, VPC and subnets through a cross-account IAM role.",
        "lower-cost": "Sprinters charges $0.01 per job and AWS bills the compute, with spot instances supported.",
        "arm64": "Ubuntu arm64 images run on Graviton instance types such as t4g and m8gd.",
        "security": "Each job gets a new instance with no ingress, and Sprinters documents having no access to repository contents, secrets or instance logins."
      },
      "peers": [
        "runs-on",
        "hyperenv",
        "cirun",
        "terraform-aws-github-runner",
        "aws-codebuild-runners",
        "warpbuild",
        "ubicloud",
        "github-hosted-runners"
      ],
      "sources": [
        "https://sprinters.sh/",
        "https://sprinters.sh/pricing",
        "https://sprinters.sh/docs",
        "https://sprinters.sh/docs/setup",
        "https://sprinters.sh/docs/label",
        "https://sprinters.sh/docs/images",
        "https://sprinters.sh/docs/instances",
        "https://sprinters.sh/docs/temp",
        "https://sprinters.sh/docs/aws-regions",
        "https://sprinters.sh/docs/security",
        "https://sprinters.sh/new",
        "https://sprinters.sh/impressum",
        "https://github.com/sprinters-sh/sprinters"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "hyperenv",
      "name": "HyperEnv for GitHub Actions Runner",
      "company": "widdix GmbH",
      "hq": {
        "city": "Lichtenwald",
        "region": "Baden-Württemberg",
        "country": "DE",
        "source": "https://hyperenv.com/imprint/"
      },
      "kind": "byoc",
      "category": "your-cloud",
      "website": "https://hyperenv.com",
      "docs": "https://hyperenv.com/help/",
      "pricing_url": "https://hyperenv.com/pricing/",
      "github": null,
      "tagline": "AWS Marketplace product that runs each GitHub Actions job on an ephemeral EC2 instance, deployed entirely inside your own AWS account.",
      "ci_systems": [
        "GitHub Actions"
      ],
      "os": [
        "Linux"
      ],
      "arch": [
        "x64"
      ],
      "isolation": "Each job gets a new EC2 instance that registers an ephemeral GitHub runner and is terminated after the job completes or times out",
      "isolation_short": "EC2 instance",
      "isolation_class": "vm",
      "ephemeral": true,
      "open_source": false,
      "license": null,
      "self_host": true,
      "byoc": true,
      "clouds": [
        "AWS"
      ],
      "gpu": true,
      "gpu_detail": "g4dn instance types (xlarge to 16xlarge) with an NVIDIA T4 GPU are supported; the HyperEnv fee stays $0.002 per build minute plus the AWS instance cost.",
      "hardware": "A fixed list of EC2 instance families: t3, m5, m6i, m7i, m7i-flex, m7a and g4dn.",
      "performance": null,
      "sizes": "Up to 192 vCPUs, 768 GiB memory and a 16 TB disk volume; the default instance type is m6i.large",
      "max_size_short": "192 vCPU / 768 GB",
      "cache": "An S3-backed cache in your account is enabled by default (S3Cache parameter, 30-day expiration); it is not supported for GitHub Enterprise Server.",
      "docker_layer_cache": null,
      "concurrency": null,
      "job_limits": "GitHubJobTimeout defaults to 120 minutes and is a configurable stack parameter",
      "static_ips": null,
      "ssh_access": true,
      "observability": "Runner logs in CloudWatch Logs, a Step Functions execution per job, VPC flow logs and optional infrastructure alarm emails.",
      "setup": "Subscribe on AWS Marketplace, launch the CloudFormation template (dedicated or existing VPC), install the private GitHub App from the stack output, then use the hyperenv label in runs-on.",
      "runs_on": "hyperenv",
      "pricing": "$0.002 per build minute, billed through AWS Marketplace and charged only while a job runs. AWS infrastructure such as EC2, EBS and CloudWatch is billed separately to your account.",
      "price_short": "$0.002/min fee + AWS cost",
      "free_tier": null,
      "compliance": "ISO 27001 (widdix GmbH)",
      "regions": null,
      "best_for": "Teams that buy through AWS Marketplace and want GitHub Actions runners that run entirely inside their AWS account for a metered per-minute fee",
      "overview": [
        "**HyperEnv for GitHub Actions Runner** is sold by widdix GmbH on AWS Marketplace as an AMI plus CloudFormation templates. The stack deploys an API Gateway endpoint, Lambda functions, Step Functions state machines and DynamoDB tables in your account. A GitHub webhook starts a state machine that launches one EC2 instance per job, trying spot in several zones before falling back to on-demand, and terminates it afterwards. The vendor describes it as having no external dependencies.",
        "Runners are based on Ubuntu 22.04 with the same packages as GitHub-hosted runners, per the FAQ. Defaults are set as stack parameters, and a job can override instance type and spot use with a `hyperenv-config` label. The product supports GitHub Free, Pro and Enterprise as well as GitHub Enterprise Server, an S3 cache and SSH access with a key pair. The fee is [$0.002 per build minute](https://hyperenv.com/pricing/) through AWS Marketplace, plus AWS costs."
      ],
      "notable": [
        "Release 3.0.0 (16 October 2025) moved runner orchestration to AWS Step Functions.",
        "An S3 cache for self-hosted workflows was announced on 20 May 2025 for HyperEnv 2.14.0 and later.",
        "v3.6.0 (18 June 2026) added support for shared (cross-account) subnets.",
        "The latest release listed is v3.7.1 (29 October 2026), updating to GitHub Runner 2.338.2."
      ],
      "pros": [
        "You want the runner control plane and every runner inside your own AWS account with no external dependencies",
        "You prefer to buy and pay through AWS Marketplace on a metered per-build-minute rate",
        "You need runners with access to private networks in an existing VPC",
        "You use GitHub Enterprise Server, which the product supports",
        "You need NVIDIA T4 GPU instances for some jobs"
      ],
      "cons": [
        "Runners are based on Ubuntu 22.04 only, per the FAQ; other operating systems are by request",
        "Instance types are limited to a fixed list of x64 families plus g4dn",
        "Per-job overrides cover only instance type and spot; other settings are stack-wide parameters",
        "The S3 cache is not supported for GitHub Enterprise Server"
      ],
      "verdict": "HyperEnv is a CloudFormation-deployed runner system for GitHub Actions that lives entirely in your AWS account, suited to teams that want AWS Marketplace billing and a small set of stack-level settings.",
      "use_cases": {
        "byoc": "All components and runners are deployed by CloudFormation into your own AWS account.",
        "lower-cost": "The fee is $0.002 per build minute and spot instances with on-demand fallback are supported.",
        "gpu": "g4dn instance types with an NVIDIA T4 GPU can be selected per job.",
        "security": "Each job runs on a single-use instance inside your VPC, and the vendor states ISO 27001 certification."
      },
      "peers": [
        "runs-on",
        "sprinters",
        "cirun",
        "terraform-aws-github-runner",
        "aws-codebuild-runners",
        "warpbuild",
        "actions-runner-controller"
      ],
      "sources": [
        "https://hyperenv.com/",
        "https://hyperenv.com/pricing/",
        "https://aws.amazon.com/marketplace/pp/prodview-7bzr422ymm2hw",
        "https://hyperenv.com/help/",
        "https://hyperenv.com/help/overview/",
        "https://hyperenv.com/help/configuration/",
        "https://hyperenv.com/help/faq/",
        "https://hyperenv.com/help/release-notes/",
        "https://hyperenv.com/setup-guide-step-01/",
        "https://hyperenv.com/blog/",
        "https://hyperenv.com/imprint/"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "aspect-workflows",
      "name": "Aspect Workflows",
      "company": "Aspect Build Systems, Inc.",
      "hq": null,
      "kind": "byoc",
      "category": "your-cloud",
      "website": "https://aspect.build",
      "docs": "https://aspect.build/docs/aspect-workflows",
      "pricing_url": "https://aspect.build/pricing",
      "github": null,
      "tagline": "Bazel-focused platform whose auto-scaling CI runners register with your existing CI and can run in your own AWS account or GCP project.",
      "ci_systems": [
        "GitHub Actions",
        "Buildkite",
        "GitLab CI",
        "CircleCI"
      ],
      "os": [
        "Linux"
      ],
      "arch": [
        "x64",
        "arm64"
      ],
      "isolation": "Persistent Linux VMs in runner groups take jobs one after another so Bazel state stays warm; isolation between jobs is not documented",
      "isolation_short": "Persistent VM",
      "isolation_class": "vm",
      "ephemeral": false,
      "open_source": false,
      "license": null,
      "self_host": true,
      "byoc": true,
      "clouds": [
        "AWS",
        "Google Cloud"
      ],
      "gpu": true,
      "gpu_detail": "GPU runner groups can use GPU instance types offered by AWS or GCP; no GPU models or prices are published.",
      "hardware": "Any AWS or GCP instance type per runner group, with local NVMe for the Bazel output base.",
      "performance": "Aspect says cached builds routinely finish in under a minute on its runners (vendor claim)",
      "sizes": "Any AWS or GCP instance type, configured per runner group",
      "max_size_short": null,
      "cache": "Runners keep the Bazel output base on NVMe between jobs and restore a warming archive on boot. The deployment includes a remote cache backed by S3 Express One Zone on AWS or Cloud Storage on GCP.",
      "docker_layer_cache": null,
      "concurrency": "Set per runner group through minimum size, maximum size and idle runners; the Enterprise fee is priced by a vCPU ceiling",
      "job_limits": null,
      "static_ips": null,
      "ssh_access": null,
      "observability": "A Build Results UI with build and test results, logs, timing profiles, flaky-test tracking and build comparison, plus an API and MCP server.",
      "setup": "Aspect Enterprise is deployed with Terraform into your AWS account or GCP project (or hosted by Aspect); jobs then select a runner group by runs-on label, Buildkite queue, GitLab tag or CircleCI resource class.",
      "runs_on": null,
      "pricing": "CI runners are part of Aspect Enterprise, which has an annual platform fee based on the vCPU ceiling set for remote execution and CI runners, the deployment shape and the support level; no amount is published. When self-hosted, compute is billed by your cloud provider.",
      "price_short": "Annual platform fee (quote) + cloud cost",
      "free_tier": "30-day trial of Aspect Enterprise",
      "compliance": "SOC 2 Type 2",
      "regions": "Your own AWS account or GCP project, including AWS GovCloud and GCP Assured Workloads",
      "best_for": "Bazel monorepo teams that want warm, persistent CI runners in their own AWS or GCP account alongside remote cache and remote execution",
      "overview": [
        "**Aspect Workflows** is a platform for Bazel from Aspect Build Systems, Inc. that combines a remote cache, remote execution, a Build Results UI and CI runners. The runners are auto-scaling Linux VMs that register with GitHub Actions, Buildkite, GitLab or CircleCI as that system’s self-hosted runners. They are persistent and NVMe-backed: jobs run one after another on a machine so the Bazel output base and analysis cache stay warm.",
        "CI runners are available on **Aspect Enterprise**, an isolated deployment either hosted by Aspect or self-hosted in your AWS account or GCP project with Terraform. Runner groups define instance type, minimum and maximum size, idle runners and a warming archive, and scale on queue depth down to zero. When self-hosted, you choose how much access Aspect holds, from fully managed to none. Pricing is an [annual platform fee](https://aspect.build/pricing) that is not published."
      ],
      "notable": [
        "The Cloud Service Terms were last updated on 20 January 2026.",
        "The pricing page (checked 2026-11-22) says Linux runners on the multi-tenant Aspect Cloud are slated for early 2027.",
        "macOS runners, OCI image-based runners and an on-prem Helm deployment are listed as coming soon (checked 2026-11-22)."
      ],
      "pros": [
        "You build with Bazel and want runners that keep the output base and analysis cache warm between jobs",
        "You want the runners, remote cache and remote execution on one private network in your own AWS account or GCP project",
        "You use Buildkite, GitLab or CircleCI rather than only GitHub Actions",
        "You need air-gapped, AWS GovCloud or GCP Assured Workloads deployments",
        "You want a vendor that can operate the deployment for you at an access level you choose"
      ],
      "cons": [
        "Runners are persistent and take jobs one after another rather than being single-use",
        "CI runners are available only on Aspect Enterprise, not on Aspect Cloud",
        "No price is published: Enterprise is an annual platform fee based on configuration",
        "macOS runners are listed as coming soon, and Azure and Oracle Cloud as planned"
      ],
      "verdict": "Aspect Workflows is a Bazel build platform whose persistent, auto-scaling runners can run in your own AWS or GCP account, suited to Bazel monorepo teams on GitHub Actions, Buildkite, GitLab or CircleCI.",
      "use_cases": {
        "byoc": "Aspect Enterprise can be self-hosted with Terraform in your own AWS account or GCP project.",
        "faster-builds": "Persistent NVMe-backed runners and warming archives keep Bazel state warm between jobs, which Aspect says shortens cached builds.",
        "gitlab": "Runner groups register with GitLab and are selected with GitLab CI tags.",
        "gpu": "GPU runner groups are available for model training and inference tests.",
        "arm64": "Runner groups can use Arm instance types alongside x86 groups."
      },
      "peers": [
        "buildkite-elastic-ci-stack",
        "runs-on",
        "actions-runner-controller",
        "gitlab-runner",
        "terraform-aws-github-runner",
        "buildkite-hosted-agents",
        "namespace",
        "depot"
      ],
      "sources": [
        "https://aspect.build/ci-runners",
        "https://aspect.build/pricing",
        "https://aspect.build/docs/aspect-workflows/get-started/overview",
        "https://aspect.build/docs/aspect-workflows/features/ci-runners",
        "https://aspect.build/docs/aspect-workflows/enterprise/ci-runners.md",
        "https://aspect.build/docs/aspect-workflows/enterprise/overview.md",
        "https://aspect.build/docs/aspect-workflows/enterprise/deployment-options/self-hosted.md",
        "https://aspect.build/docs/aspect-workflows/enterprise/trial.md",
        "https://aspect.build/security",
        "https://aspect.build/terms"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    },
    {
      "id": "enomem",
      "name": "enomem",
      "company": "enomem",
      "hq": null,
      "kind": "byoc",
      "category": "your-cloud",
      "website": "https://enomem.dev",
      "docs": "https://enomem.dev/setup/",
      "pricing_url": "https://enomem.dev/#pricing",
      "github": null,
      "tagline": "Self-hosted GitHub Actions runners on Azure: an Azure Functions handler in your subscription starts one Spot VM per job.",
      "ci_systems": [
        "GitHub Actions"
      ],
      "os": [
        "Linux"
      ],
      "arch": [
        "x64",
        "arm64"
      ],
      "isolation": "Each job gets a fresh Spot VM with an ephemeral OS disk in its own resource group, which is deleted when the job completes",
      "isolation_short": "Azure VM",
      "isolation_class": "vm",
      "ephemeral": true,
      "open_source": null,
      "license": null,
      "self_host": true,
      "byoc": true,
      "clouds": [
        "Azure"
      ],
      "gpu": null,
      "gpu_detail": null,
      "hardware": "Azure D-, E- and F-series sizes that support ephemeral OS disks, on AMD x64 and Cobalt ARM64.",
      "performance": null,
      "sizes": "Built-in aliases from 2 to 32 vCPU on x64 and 2 to 16 vCPU on ARM64 (D-series, 4 GB per vCPU); any Standard_* size with ephemeral OS disk support can be used, up to 96 vCPU / 384 GB",
      "max_size_short": "96 vCPU / 384 GB",
      "cache": "A cache sidecar redirects actions/cache to Azure Blob Storage in your region, shared across the organisation’s repositories. Docker layer caching is documented through a pre-configured Buildx builder with cache-from and cache-to.",
      "docker_layer_cache": true,
      "concurrency": "Team: up to 50 concurrent runners; Enterprise: unlimited concurrent runners",
      "job_limits": "Resource groups older than 4 hours are deleted by a cleanup timer",
      "static_ips": null,
      "ssh_access": null,
      "observability": "Application Insights logging and monitoring is deployed with the stack.",
      "setup": "Install the enomem CLI, activate a licence key, run enomem setup to create a GitHub App and deploy the Azure resources, then set runs-on: enomem in your workflows.",
      "runs_on": "enomem",
      "pricing": "Free: $0 for personal projects, open source and non-commercial use. Team: $25/month or $300/year, up to 50 concurrent runners. Enterprise: $125/month or $1,500/year, with pre-built VM images and unlimited concurrent runners. Azure Spot VM costs are paid directly to Azure.",
      "price_short": "$25/mo licence + Azure cost",
      "free_tier": "Free licence for personal projects, open source and non-commercial use, with unlimited runners",
      "compliance": null,
      "regions": "Any Azure region that supports the chosen VM sizes, selected at setup",
      "best_for": "Teams on GitHub Actions that already use Azure and want per-job Spot VMs inside their own subscription for a flat monthly licence",
      "overview": [
        "**enomem** runs GitHub Actions jobs on Azure Spot VMs in your own subscription. Its `enomem setup` CLI creates a GitHub App and deploys an Azure Functions app, Key Vault, storage account, VNet and managed identity. When GitHub sends a queued-job webhook, the function creates a resource group with one VM that has an ephemeral OS disk and no inbound access, the runner takes exactly one job, and the resource group is deleted afterwards.",
        "Jobs pick a size with a label alias such as `enomem-8vcpu`, with overrides for VM size, spot and cache. Runners use Ubuntu 24.04 LTS on x64 or ARM64, and a timer re-provisions on-demand when a spot deployment fails or stalls. The Team tier boots a stock image with cloud-init, while Enterprise adds pre-built images. Licences are [flat monthly fees](https://enomem.dev/#pricing) and Azure bills the VMs. The site does not name the company behind the product."
      ],
      "notable": [
        "The three posts on the enomem blog are all dated 28 April 2026.",
        "Runner images are rebuilt biweekly, according to the configuration page (checked 2026-11-22).",
        "The site footer carries a 2026 copyright and no launch announcement is published (checked 2026-11-22)."
      ],
      "pros": [
        "You run on Azure and want GitHub Actions runners inside your own subscription, VNet and region",
        "You want each job on a single-use Spot VM in its own resource group, with on-demand as a fallback",
        "You prefer a flat monthly licence, or qualify for the free non-commercial licence",
        "You need ARM64 runners on Azure Cobalt sizes",
        "You want actions/cache and Docker layer caches stored in your own Azure Blob Storage"
      ],
      "cons": [
        "Windows and macOS are not currently supported",
        "On the Team tier a stock Ubuntu image is used and cloud-init takes 45 to 55 seconds before steps start; pre-built images are Enterprise only",
        "The Team tier is limited to 50 concurrent runners",
        "The operating company, its location and terms of service are not documented on the site"
      ],
      "verdict": "enomem is licensed software that runs single-use GitHub Actions runners on Azure Spot VMs in your own subscription, suited to Azure-based teams that want a flat monthly fee.",
      "use_cases": {
        "byoc": "The webhook handler, cache and every runner VM are deployed in your own Azure subscription.",
        "lower-cost": "Runners use Azure Spot VMs billed per second by Azure, with a flat monthly licence from enomem.",
        "arm64": "ARM64 aliases run on Azure Cobalt D-series sizes from 2 to 16 vCPU.",
        "docker-builds": "A pre-configured Buildx builder supports Docker layer caching with cache-from and cache-to."
      },
      "peers": [
        "azure-managed-devops-pools",
        "runs-on",
        "cirun",
        "sprinters",
        "hyperenv",
        "warpbuild",
        "actions-runner-controller",
        "garm"
      ],
      "sources": [
        "https://enomem.dev/",
        "https://enomem.dev/setup/",
        "https://enomem.dev/architecture/",
        "https://enomem.dev/configuration/",
        "https://enomem.dev/faq/",
        "https://enomem.dev/compare/",
        "https://enomem.dev/blog/"
      ],
      "verified_on": "2026-11-22",
      "alternatives": true
    }
  ]
}