Skip to main content
Orbi turns a GitHub Issue into a tested, independently reviewed, and merged pull request. Orbi is a local AI development worker. You put work into GitHub Issues; it automatically claims an ai-ready Issue, develops it in an isolated worktree, runs the real test suite, opens a PR, then completes an independent review (with in-session fixes) and merges a clean PR — with the whole delivery recorded in GitHub Issues, comments and the PR itself. The reviewing session can run on a different model than the implementing session (learn how). This MVP keeps state only in GitHub rather than a database or queue, runs one Issue per tick rather than a daemon, and does not discover repositories or push to protected branches itself. See /security for the full boundary list.

What makes it different

GitHub is the whole ledger

Issues, labels, comments and the PR are the complete delivery record. Orbi uses no database or second task system.

Independent review, optionally on a different model

A second session that did not write the code reviews the PR and fixes findings in-session. review_pi_* lets that session run on another model.

You can steer a running delivery

A trusted correction on the Issue restarts the run with it.

Releases are deliveries too

A Release task runs the release state machine: CI wait, tag, and notes. Release work follows the same GitHub delivery record.

Self-hosted core, hosted control plane

The runner is open source and runs on your machine. Orbi Managed Cloud runs the same ledger without you operating a runner.

How it works

Where to go next

  • Quickstart: create and deliver your first Issue.
  • Ways to run Orbi: self-host the runner or use Managed Cloud.
  • Workflow: the Issue → PR → review → merge chain.
  • Operations: timers, logs, CLI, worktrees and recovery.
  • Security: delivery boundaries and protected-branch safeguards.