# SuperCI

> Free, open-source GitHub Actions runners on AWS spot instances in your own account: a fresh machine per job at about an eighth of GitHub's price.

Start the dashboard with `npx @superci/cli dashboard`. A workflow runs on SuperCI when it says `runs-on: superci`. Everything the dashboard does is also a command: `npx @superci/cli` lists them, each takes `--json`, a change takes `--dry-run`, and none asks a question.

Source: https://github.com/supercorp-ai/superci

Every page of this site is also Markdown: add `.md` to its address, or ask for it with `Accept: text/markdown`. All the guides in one file: https://superci.dev/llms-full.txt

## Documentation

- [Getting started](https://superci.dev/docs/getting-started.md): Start the dashboard, put a control plane in your cloud, connect a repository and run your first job on your own machines.
- [How it works](https://superci.dev/docs/how-it-works.md): The three parts of SuperCI (a dashboard, a control plane in your cloud, a machine per job) and what happens from a push to a finished job.
- [Commands and coding agents](https://superci.dev/docs/commands.md): Run everything the dashboard does from a terminal or a coding agent: sign in once, then status, runners, repositories and updates as commands with JSON output.
- [Workspaces](https://superci.dev/docs/workspaces.md): Look after several CIs from one computer: each workspace has its own sign-ins, control plane and runners, a switcher in the dashboard, and --workspace in commands.
- [Labels and machines](https://superci.dev/docs/labels.md): How the runs-on label picks a job's machine, as key=value pairs after its name: size, architecture, GPUs, Windows, the disk, a provider, or no spot machines.
- [Runner providers](https://superci.dev/docs/providers.md): What AWS, Cloudflare and Modal each run, what SuperCI creates in each account, and what each one needs from you.
- [What your AWS account allows](https://superci.dev/docs/aws-accounts.md): How SuperCI looks at an AWS account and works with what it allows: its regions, spot machines, CPUs at once and the runner image, and a project from AWS's new sign-up.
- [Order and limits](https://superci.dev/docs/order-and-limits.md): How jobs are placed across your providers, what happens when one cannot start a job, and how to cap jobs at once and spending by month and by day, in SuperCI and in AWS itself.
- [Several accounts](https://superci.dev/docs/several-accounts.md): Run one CI's jobs in several AWS accounts, Cloudflare accounts or Modal workspaces, each with its own limits and spend
- [Spot machines](https://superci.dev/docs/spot-machines.md): How SuperCI uses AWS spot machines, what happens when AWS takes one back mid-job, and how to keep a job off spot.
- [Runs from a computer](https://superci.dev/docs/runs-from-a-computer.md): Run a workflow's jobs on your own machines from your computer or a coding agent, with the working tree as it is and nothing pushed.
- [Cache](https://superci.dev/docs/cache.md): GitHub's cache, and Turbo's, Nx's, Bazel's and GitLab's, kept in a bucket in your own AWS account: who may read what, what it keeps, and how to set it.
- [Faster tests](https://superci.dev/docs/faster-tests.md): Make a test job finish sooner on SuperCI: a bigger machine with every CPU in use, a matrix for tests that cannot share one, and less setup, all measured.
- [GPU runners](https://superci.dev/docs/gpu.md): Run GitHub Actions jobs on an NVIDIA GPU in your own AWS or Modal account: the labels, what each GPU costs, and the quota AWS starts you at.
- [Windows runners](https://superci.dev/docs/windows.md): Run GitHub Actions jobs on Windows Server 2025 machines in your own AWS account: the label, the sizes, prices beside GitHub's, and what differs from Linux.
- [macOS runners](https://superci.dev/docs/macos.md): Coming soon: macOS jobs on Apple Macs in your own AWS account, each Mac rented for 24 hours, reused with a clean disk for every job.
- [Arm runners](https://superci.dev/docs/arm64.md): Run GitHub Actions and GitLab CI jobs on Arm (AWS Graviton) machines in your own account: the label, the prices, and building for two architectures.
- [Larger runners](https://superci.dev/docs/larger-runners.md): Move jobs from GitHub's larger runners to machines of the same size in your own cloud: which label replaces which runner, what each costs, and what changes.
- [GitHub](https://superci.dev/docs/github.md): Connect an organization or your own account on GitHub: the GitHub App SuperCI creates for you, what it may do, choosing repositories, and what to check when a job never starts.
- [GitLab CI](https://superci.dev/docs/gitlab.md): Connect GitLab.com or a self-managed GitLab with an access token, turn projects on, and run jobs by tag.
- [GitHub Enterprise](https://superci.dev/docs/github-enterprise.md): Use SuperCI with GitHub Enterprise Server or GitHub Enterprise Cloud with data residency, and what your server has to allow.
- [Public repositories](https://superci.dev/docs/public-repositories.md): Why jobs from public repositories are refused by default, how to allow one, and which runs are refused even then.
- [Your own network on AWS](https://superci.dev/docs/private-networks.md): Start AWS machines in your own subnets and security groups, so jobs reach what that network reaches.
- [Troubleshooting](https://superci.dev/docs/troubleshooting.md): What to check when a job waits, fails at once or is refused, and where SuperCI says why.

## Compare

- [Blacksmith alternatives for GitHub Actions (2026)](https://superci.dev/compare/blacksmith-alternatives.md): Alternatives to Blacksmith's GitHub Actions runners: Depot, Namespace, WarpBuild, Ubicloud, RunsOn, SuperCI, GitHub and ARC, with their prices per minute.
- [BuildJet alternatives for GitHub Actions (2026)](https://superci.dev/compare/buildjet-alternatives.md): BuildJet stopped running jobs on 31 March 2026. Where its users can go: GitHub, Blacksmith, Depot, Namespace, WarpBuild, Ubicloud, RunsOn or SuperCI.
- [WarpBuild alternatives for GitHub Actions (2026)](https://superci.dev/compare/warpbuild-alternatives.md): Alternatives to WarpBuild's hosted runners and to WarpBuild BYOC: Blacksmith, Namespace, Ubicloud, RunsOn, SuperCI and GitHub, with prices per minute.
- [Self-hosted GitHub Actions runner pricing (2026)](https://superci.dev/compare/self-hosted-runner-pricing.md): GitHub charges nothing for self-hosted runners; its $0.002 fee was postponed. What they cost on your machines, Kubernetes, spot, RunsOn, WarpBuild or SuperCI.
- [SuperCI vs GitHub-hosted runners](https://superci.dev/compare/github-hosted-runners.md): GitHub's own runners need no set-up and are free for public repositories. When a spot machine in your own cloud is the better choice, and how to use both.
- [SuperCI vs Depot, Blacksmith and Namespace](https://superci.dev/compare/hosted-runner-services.md): Depot, Blacksmith and Namespace run your GitHub Actions jobs on their machines. SuperCI starts machines in your own cloud. Prices, speed, caches and trust.
- [SuperCI vs RunsOn: runners in your own cloud](https://superci.dev/compare/runs-on.md): SuperCI and RunsOn both run GitHub Actions jobs on machines in your own AWS account. How they differ in clouds, code hosts, licence, set-up and limits.
- [SuperCI vs Actions Runner Controller (ARC)](https://superci.dev/compare/actions-runner-controller.md): Actions Runner Controller runs GitHub Actions runners as pods in a Kubernetes cluster you operate. SuperCI starts a machine per job, with no cluster.
- [SuperCI vs the Terraform module for runners on AWS](https://superci.dev/compare/terraform-aws-github-runner.md): terraform-aws-github-runner scales GitHub Actions runners on EC2 with Terraform and Lambda. How it compares with SuperCI in set-up, clouds and upkeep.

## Blog

- [GitHub Actions limits: time, memory, disk, cache and more](https://superci.dev/blog/github-actions-limits.md): Every GitHub Actions limit in one place: job time, memory, disk, jobs at once, the matrix, cache, storage, the API and spending, and which can be raised.
- [GitHub Actions budget is preventing further use: what to do](https://superci.dev/blog/github-actions-budget-preventing-further-use.md): Why GitHub says "The job was not started because an Actions budget is preventing further use", which budget it is, and how to get jobs running again.
- [GitHub Actions out of memory: exit code 137 and heap errors](https://superci.dev/blog/github-actions-out-of-memory.md): Why GitHub Actions jobs run out of memory: how much RAM each runner has, what exit code 137 means, and how to fix heap errors in Node, Java and test runners.
- [No space left on device in GitHub Actions: causes and fixes](https://superci.dev/blog/github-actions-no-space-left-on-device.md): Why GitHub Actions jobs run out of disk: how much space each runner really has, what fills it (Docker most of all), and five ways to get room back.
- [GitHub Actions time limit: how to run jobs past 6 hours](https://superci.dev/blog/github-actions-6-hour-limit.md): GitHub Actions' time limits: a job stops at 6 hours on GitHub's runners and timeout-minutes defaults to 360. Three ways to run longer jobs.
- [Waiting for a runner to pick up this job: what to check](https://superci.dev/blog/waiting-for-a-runner-to-pick-up-this-job.md): Why a GitHub Actions job stays at "Waiting for a runner to pick up this job": labels, plan limits, larger runners, billing, your own runners and autoscalers.
- [Self-hosted GitHub Actions runners on AWS](https://superci.dev/blog/github-actions-runners-on-aws.md): Five ways to run GitHub Actions jobs in your own AWS account: a runner on EC2, ECS on Fargate, CodeBuild, ARC on EKS, or a fresh machine for each job.
- [What a GitHub Actions minute costs in 2026](https://superci.dev/blog/github-actions-pricing.md): GitHub Actions prices by runner size after the January 2026 cut, how minutes are counted, what happened to the self-hosted charge, and ways to pay less.
- [Self-hosted GitHub Actions runners without Kubernetes](https://superci.dev/blog/self-hosted-runners-without-kubernetes.md): GitHub's own way to scale self-hosted runners needs a Kubernetes cluster. What one fresh machine per job takes without one, and four ways to get it.
- [GitLab CI jobs on AWS spot machines](https://superci.dev/blog/gitlab-ci-aws-spot-machines.md): What a GitLab CI minute costs on GitLab.com and on an AWS spot machine, how GitLab's own autoscaler does it, and a way with no runner manager to keep.
- [GitHub Actions jobs in Cloudflare containers](https://superci.dev/blog/github-actions-cloudflare-containers.md): What a GitHub Actions job costs in a Cloudflare container, which jobs fit in 4 CPUs and 12 GiB and which do not, and how to send jobs there.
- [GitHub Actions on spot instances: an eighth of the price](https://superci.dev/blog/cheaper-github-actions-runners-aws-spot.md): What a GitHub Actions minute costs on GitHub's runners and on AWS spot instances, where the saving comes from, what you give up, and how to set it up.

## Prices

- [GitHub Actions cost calculator](https://superci.dev/calculator.md): Enter minutes and pick a machine: a month of GitHub Actions jobs on GitHub's runners, and on the same size in your own AWS, Cloudflare or Modal account.

## Optional

- [Changelog](https://superci.dev/changelog.md): What changed in each release.
- [Security](https://superci.dev/security.md): How SuperCI is built, where its trust lines are, and how to report a problem.
