AUTOP

How Autop works

Autop watches your GitHub Projects board, dispatches each ready issue to a runner on your machine, where Claude Code or Codex implements it, your checks run and an automated review posts its verdict. Merging follows the autonomy level you set per project. Autop coordinates; it never runs models or clones repositories itself.

What happens to an issue?

  1. Issue: you file it on your board with the autop label and a priority.
  2. Dispatch: Autop leases it to a runner with capacity.
  3. Build: your coding agent implements it in a worktree on your machine.
  4. Check: your build, tests and lint run on the same machine.
  5. Pull request: the runner publishes the branch and opens the PR.
  6. Review: an automated review posts its verdict; hands-on projects wait for you.
  7. Merge: by your branch rules and autonomy level; deployment stays in your pipeline.

Where does everything live?

GitHub carries the issues and pull requests. The Autop plane schedules work and enforces your rules. Your machine runs the coding agent, the checks, and holds every credential.

Your GitHub organisationAutopYour machine
Your code and worktreesRepositoriesnoClones and worktrees
Your coding agent and its loginnonoyes (never read or copied)
Issues, pull requests, reviewsyesPosts verdicts, merges by your rulesPublishes branches and PRs
Scheduling, leases, your autonomy rulesBoard statusyesno
Model callsnono (never)Your agent to its provider
Publishing tokennoMinted per jobUsed, never stored

Deployment stays with your own pipeline.

How often does Autop ask you?

Set the level per project. Your project settings decide; nothing on this page changes them.

Which coding agents and machines?

Claude Code and Codex, installed and logged in on a Mac or Linux machine that stays on: a Mac mini, a workstation, or a server you rent. Autop-hosted machines are planned.

Join the waitlist