Orbi

Orbi is an open-source software delivery system for GitHub, developed under the orbi-build organization and operated by Low Code, LLC. You label an Issue ai-ready, and Orbi implements it in an isolated worktree, opens a pull request, takes it through independent review, merges it, and can run release gates to publish a tag and release notes. It can be self-hosted or used through a managed Cloud offering.

What It Is

Orbi describes itself as a "software dark factory" or lights-out software factory. GitHub Issues act as the task pool and the only state store, so there is no separate database, queue, or daemon. Each task gets its own run, with a branch, worktree, logs, and PR tied to the same run ID.

How the delivery loop works

A scheduler timer (systemd on Linux, launchd on macOS) triggers a tick every 5 minutes. Orbi claims an ai-ready Issue, creates a feature branch and worktree from a frozen base SHA, and has the Pi coding agent plan, implement, test, and verify. The runner then syncs the base, pushes, and opens a PR with Fixes #N.

A separate review session checks the exact diff against the Issue's acceptance criteria, can fix findings in the same session, and reruns the tests. Only the reviewed head may merge. Unrecoverable failures mark the Issue ai-blocked for a human decision. Restarts reconstruct state from the Issue, PR, labels, journal, and run ID.

Setup and models

Self-hosting needs Python 3.14 or newer, git, the GitHub CLI, a scheduler session, and access to a model. Orbi works with OpenAI-compatible APIs or locally hosted models. The CLI offers setup, doctor, add, status, session, and install-units commands, and is also published on PyPI as orbi-cli.

Licensing and Cloud

The core is licensed under AGPL-3.0, with the Sustainable Use License as an alternative. A managed Cloud option starts from GitHub sign-in and App installation, and the vendor says its runner clones authorized repositories onto machines it operates during a delivery.