1on: push2jobs:3 build:- runs-on: ubuntu-latest+ runs-on:6 steps:7 - uses: actions/checkout@v48 - run: pytest
- Build GPU / build (3.7) (pull_request)Queued
- Windows / win_build (3.7) (pull_request)Queued
- Build / build (3.8) (pull_request)Queued
- Build GPU / build (3.8) (pull_request)Queued
- Windows / win_build (3.8) (pull_request)Queued
- Cirun— Cirun.io: Managed self-hosted GitHub Actions runners on your cloud.
Native
It's native to GitHub — easily see checks and logs in GitHub's interface.
Fast
Super-fast CI with more performant machines from your favourite cloud provider.
Compare your CI costs
Many managed GitHub Actions providers bill per minute. Cirun runs on your own cloud, so the compute bill is whatever your cloud charges. Plug in your numbers to see how it shakes out for your workload.
Per-minute pricing reflects typical managed GitHub Actions providers (matched core count, ~$0.002/core/min). Cloud on-demand prices for Linux: Azure D-ads v5 (PAYG), AWS m7a (us-east-1), GCP n2d-standard (us-central1) — sourced from each cloud’s published pricing via Vantage. Spot / preemptible factors derived from Vantage’s spot rates at the 16-vCPU tier of each family (Azure ~80% off, AWS ~49% off, GCP ~79% off); real spot prices fluctuate by region, AZ, family and demand.
1runners:2 - name: gpu-runner3 # Cloud Provider: GCP4 cloud: gcp5 # Cheapest GPU on GCP6 gpu: nvidia-tesla-t47 # Cheapest VM on GCP, with GPU attachable8 instance_type: n1-standard-19 # Custom image with NVIDIA drivers installed on Ubuntu-20.410 machine_image: sgkit-dev/cirun-nvidia11 # preemptible instances seems quite less reliable.12 preemptible: false13 # Path of the relevant workflow file14 labels:15 - cirun-gcp-gpu-runner
Easy to Configure
Runners can be configured with a one-line change; advanced setups stay in a simple YAML file.

On-Demand Runners
Only pay your cloud provider for what you use. Runners are terminated immediately after job completion.

Unlimited Customization
Use a custom image or pre-existing images from your cloud provider. (Example: NVIDIA AMI from AWS Marketplace.)
Free for Open Source
We ♥️ Open Source. Cirun.io is free for Open source projects.
GPU Support
We also support GPU enabled runners on your cloud provider.
Preemptible runners
We support low cost preemptible instances on GCP and AWS.
Trusted by teams running serious CI







Pricing
Pay only for what you use. Always free for open source.
- Install cirun on unlimited public repositories
- Unlimited runners
- ARM-based runners
- GPU runners
- All supported cloud providers
- Everything in Open Source
- Install cirun on 3 private repositories
- Support via Slack / email
- Everything in Open Source
- Install cirun on 10 private repositories
- Support via Slack / email
- Everything in Open Source
- Install cirun on unlimited private repositories
- Premium support via Slack / email / phone
- Contractual SLA on uptime and response times
- Custom requirements
See it in action
A one-minute walkthrough of creating on-demand, self-hosted runners on your own cloud.
What teams are saying
Using Cirun with AxisNow has been a game-changer for us. We've successfully resolved our ARM64 GitHub Action build and test issues, which will help us deliver more architecture and OS platform artifacts more gracefully and economically. The Cirun code and product are incredibly stable and easy to use. Thanks, team, for this fantastic experience!

At The Browser Company, we've always known that we need to build Arc on both Mac and Windows, in equal measure. Making Arc available to Windows 10, 11 and ARM64 users was one of our most requested features. Cirun allowed us to build a solid product in record time, cost-effectively, and at the highest quality standard from day one.

At conda-forge we rely on Cirun to deploy self-hosted runners on demand. Amit helped us immensely to integrate with an OpenStack server, and also accommodated access control rules so the whole conda-forge ecosystem could benefit from the custom runners. It now powers the Linux builds for several popular packages like Tensorflow or Pytorch, including GPU support!

Cirun was super easy to set up, works reliably, and allowed us to run on cheap large AWS spot instances which was necessary for our project's test suite that loads and tests large ML models.

We got up and running with cirun in less than an hour. It's enabled us to run GPU-backed integration tests and shaved 40% off our CI times. cirun has allowed for extreme flexibility as to which test runners we use. If you do happen to need support, their team is extremely helpful and responsive. Easily worth the small price!

We use cirun at scverse.org to easily access GPUs for our CI. It is absolutely vital for us to ensure that our code is battle-tested and cirun makes this process trivial and efficient.

The RoboStack project has been using the open source offering of cirun for the past two years. It has been remarkably stable and allowed us to build linux-aarch64 packages at an unbeatable pricepoint and speed! Very solid product!

At neurodesk.org we use Cirun for building large docker containers on our own cloud infrastructure. Using custom cloud resources for GitHub Actions lets us build these containers more cost efficiently than on the GitHub-hosted runners.

A few hours work saved us $1800 per month. Seems worth it to me.

Thank you for @CirunHQ – it's been working very well for us :)

Profitable, Bootstrapped, and Built to Last
Profitable
We're financially sustainable and growing, without relying on external funding or investors.
Independent
We make decisions based on what's best for our users and the developer community, not investors.
Long-Term Focus
Building reliable CI/CD infrastructure for the long haul, not chasing exits or acquisitions.






