Skip to main content

Introduction

Cirun is a managed self-hosted runner platform for GitHub Actions that lets you:

  • Run jobs on your own cloud — bring AWS, GCP, Azure, Hetzner, DigitalOcean, Oracle, OpenStack, or Vast.ai credentials.
  • Pay your cloud, not the runner provider — runners are cheaper than GitHub-hosted, especially at scale.
  • Pick any machine — GPUs, ARM64, big-RAM, custom images. Per-job, no fleet to manage.
  • Cache GitHub Actions without size limits — backed by your own S3 bucket.
  • Self-host on-prem — register existing hardware (Linux, macOS, Windows) as a runner pool.

You configure runners as code in a single .cirun.yml file at the root of your repo. No fleet config, no autoscaling group to manage — Cirun spins runners up on demand and tears them down when the job finishes.

Here's a minimal .cirun.yml that gives a workflow access to a 4-CPU AWS runner:

runners:
  - name: small-runner
    cloud: aws
    instance_type: c5.xlarge
    machine_image: ami-0c55b159cbfafe1f0
    region: us-east-1
    preemptible: false
    labels:
      - cirun-small

Then in .github/workflows/build.yml:

jobs:
  build:
    runs-on: cirun-small
    steps:
      - uses: actions/checkout@v4
      - run: ./build.sh

That's it — push the change and Cirun will provision a runner from your AWS account for every job tagged cirun-small.

Where to next​

  • Quickstart — connect your repo and run your first job.
  • Reference — full .cirun.yml schema and per-cloud options.
  • Cloud setup — credentials and roles for AWS, GCP, Azure, and others.
  • Caching — speed up workflows with S3-backed actions cache.
  • On-Prem — register existing hardware as runners.