Runners

Runners execute CI checks and agent jobs. Use the managed cloud pool with zero setup, or register your own machines as local runners for full control over the hardware and environment.

Cloud vs. local runners

• Cloud runners are managed by tmppr Cloud. CI dispatch and tmppr cloud run use them automatically — nothing to install, billed in runner minutes.
• Local runners are machines you enroll. They poll for queued jobs, claim the ones they are authorized for, and run them on your own hardware.

Register a local runner

tmppr runner register my-mac --kind local --labels macos,arm64
tmppr runner serve    # claim and execute queued jobs on this machine

register enrolls the machine and stores a runner credential locally; serve is the long-running daemon (use --once to process a single job).

Groups and labels

• Groups — every runner belongs to a runner group. The group's policy decides which repos it may serve: all repos in the scope, or an explicit list. Public repos are denied by default.
• Labels — a runner advertises capability labels (like macos,arm64); each job carries a selector. A job only lands on a runner whose labels cover its selector.

The security model, in plain terms

A runner can only claim a job when all of these hold:

• The job's repo is inside the runner's scope, and the runner group's policy allows that repo.
• The runner's labels cover the job's selector.
• Everyone whose input triggered the job has write access to the repo. If not, the job is held for approval — or routed only to ephemeral cloud runners with a read-only credential.
• Cross-user jobs need consent from both sides. The repo's admins must allowlist the runner's group, and the runner's owner must opt in to shared work (the default is self-only). Assignment alone never places a job on a non-consenting machine.

Minutes and pricing

Included minutes apply to managed cloud runners. Check your current usage any time with tmppr cloud status. Plans are on the pricing page.