GitHub Actions runners
Run your GitHub Actions jobs on Easel by changing runs-on, with Docker, cache volumes that mount in about a second, and per-second billing at $0.0035 per vCPU-minute.
Pro workspaces can now run GitHub Actions jobs on Easel. Connect your GitHub organization, then change runs-on in your workflow. Steps, actions, and secrets work as they do on GitHub’s runners.
Connect an organization
Open Runners in your workspace’s sidebar and select Connect organization. An owner of the organization installs the Easel Runners app and picks the repositories it covers. Each organization connects to one workspace, and that workspace pays for its jobs. Personal GitHub accounts aren’t supported, since they have no organization runners.
The Runners page lists your recent jobs with their repository, label, status, duration, and vCPU-minutes, and this billing period’s runner time and cost.


The Runners page: your connected organization, every recent job, and this period’s runner time and cost
Sizes
easel gives a job 4 vCPUs and 16 GB of memory, the same as GitHub’s default. Ask for a size with easel-1cpu, easel-2cpu, easel-4cpu, or easel-8cpu; each vCPU comes with 4 GB of memory.
Every job runs in a fresh Linux x86-64 sandbox that’s deleted when the job ends. Docker works, including service containers and docker build. A job runs for up to 5 hours.
Cache volumes
easel-sh/cache keeps directories between jobs on a disk that mounts in about a second, with no download or unpacking:
A job reads what the last successful push to your default branch left. Pull requests, including ones from forks, read that cache and never change it. Each volume holds up to 50 GiB, and one unused for 7 days is deleted. On a runner that isn’t Easel’s, the step does nothing, so the workflow still runs anywhere.
Pricing and limits
Runner time costs $0.0035 per vCPU-minute, billed per second from the sandbox’s start to its stop. A 4-vCPU job costs $0.014 a minute. It counts against Pro’s included $20 of usage like every other line, and cache volumes are included.
Each workspace runs up to 16 vCPUs of jobs at once. Jobs past that wait in GitHub’s queue and start as running ones finish.
See Pricing for the Runner Time meter.