Orbi
Self-hosted AI coding agent for GitHub Issues. Runs on Linux or macOS with local models or OpenAI-compatible APIs; free under AGPL-3.0.

Orbi is a self-hosted AI coding agent for developers and teams who want GitHub backlog tasks carried through to reviewed pull requests and tagged releases. It runs on Linux or macOS, on your own computer or server. The macOS launchd scheduler is documented but has not yet been verified on real hardware. The open source core is free under AGPL-3.0.
Its main distinction is the review stage. A separate agent session checks the proposed changes against the Issue's acceptance criteria, repairs problems, and reruns tests. Only the exact commit that passed review can merge. Existing GitHub branch protections, required checks, and human approval rules still apply, so teams keep their own gates on what reaches the repository.
Orbi works from labelled GitHub Issues and gives each task an isolated worktree. Epics and milestones define scope, while dependencies determine the order of work. GitHub holds the delivery record, including pull requests and progress comments. After an interruption, Orbi reconstructs the work from that record and continues with the same branch and pull request. Failures that need a human decision leave the Issue marked as blocked.
You can use a locally hosted model or an OpenAI-compatible API through your own credentials. No GPU is required. Self-hosting keeps the repository, keys, and test suite on your machine; a hosted model provider handles model requests when you choose that backend. The workflow needs access to GitHub.