actions-runner-controller brings Kubernetes-native autoscaling to GitHub Actions self-hosted runners, solving the cost and management challenges of maintaining a fixed fleet of CI/CD build machines. When GitHub Actions workflows queue jobs, ARC automatically provisions new runner pods on the Kubernetes cluster. When jobs complete, the pods are terminated, ensuring compute costs are proportional to actual CI/CD demand rather than peak capacity.
The controller supports both the legacy runner scale set architecture and the newer GitHub-recommended autoscaling runner scale sets. Runner groups enable organizing runners by capability, with some groups providing GPU-equipped runners for ML workloads while others provide standard compute for build and test jobs. Custom runner images allow pre-installing specific tools and dependencies to reduce job startup time.
As GitHub's official controller with over 6,100 GitHub stars, ARC has become the standard solution for organizations running GitHub Actions at scale on Kubernetes. Organization and enterprise-level runner management enables centralized control over runner resources across multiple repositories. Integration with Kubernetes resource limits and node selectors ensures runners run on appropriate hardware with bounded resource consumption.