Orbi

Orbi is an open-source AI GitHub App that turns GitHub issues into reviewed, merged pull requests. Instead of asking developers to copy context into a chat window, paste code back into an editor, and then open a pull request by hand, Orbi works where the engineering conversation already lives: the issue tracker. A maintainer writes or labels an issue, Orbi reads the issue together with the repository, plans the change, writes the code, runs the project's checks, and opens a pull request that a human can review, comment on, and merge. The goal is simple: shorten the distance between "we should fix this" and "this is fixed in main" without removing human judgment from the loop.

How it works. Orbi is installed as a GitHub App on the repositories a team chooses. When an issue is assigned to Orbi or tagged for it, the agent clones the repository into an isolated workspace, reads the relevant files, follows the conventions it finds in the codebase, and drafts an implementation. It then runs tests, linters, and build steps defined by the project so the pull request arrives with evidence rather than guesses. If reviewers leave comments on the pull request, Orbi reads them and pushes follow-up commits, the same way a teammate would respond to code review. Every change is visible as ordinary Git history, so nothing happens outside the normal review workflow and branch protections stay in force.

Who it is for. Orbi is built for small teams, open-source maintainers, indie hackers, and engineering groups that have more well-scoped issues than hands to work on them. Typical examples include bug fixes with a clear reproduction, small features, dependency upgrades, test coverage gaps, documentation corrections, refactors that touch many files in a mechanical way, and the long tail of "good first issue" tasks that never seem to get picked up. Maintainers can keep their attention on architecture, product decisions, and review, while Orbi handles the repetitive implementation work that would otherwise sit in the backlog for weeks.

Architecture. Orbi is written in Python and released under the AGPL-3.0 license, and its full source code is public at https://github.com/orbi-build/orbi. The system is composed of a GitHub App webhook receiver, a task queue that turns issue events into agent jobs, sandboxed execution environments in which the agent edits code and runs commands, and a pull request layer that pushes branches and posts summaries. Because it is open source, teams can read exactly what the agent does, audit how credentials are handled, and self-host the whole stack on their own infrastructure if they prefer to keep code and model traffic inside their own network. For teams that do not want to operate infrastructure, Orbi Cloud provides a hosted version with the same workflow.

Model flexibility. Orbi is designed to work with modern large language models for planning and code generation, and the open-source edition lets operators configure which model provider they use. This means teams can choose a model that matches their cost, latency, and privacy requirements, and can change that choice later without changing how developers interact with the tool. The interface stays the same: an issue goes in, a pull request comes out.

Safety and control. Orbi never merges on its own authority unless a repository owner explicitly configures that. By default, it opens pull requests and waits for human review. It respects branch protection rules, required status checks, and CODEOWNERS. Work happens in isolated sandboxes, and the app only receives the repository permissions that the installer grants. Because the output is a normal pull request, existing CI, security scanning, and review policies apply automatically. If a change is wrong, reviewers can request changes or simply close the pull request; nothing is hidden.

Use cases. Open-source maintainers use Orbi to clear backlogs of small, well-described issues and to respond faster to community bug reports. Startups use it to keep shipping minor fixes while senior engineers focus on core product work. Agencies use it to handle routine maintenance across many client repositories. Platform teams use it for repetitive migrations, such as renaming an API across a codebase, updating configuration formats, or adding missing type hints and tests. Documentation teams use it to fix broken links, outdated examples, and typos reported in issues. In each case, the issue tracker remains the source of truth and the pull request remains the unit of review.

Developer experience. There is no new dashboard that developers must live in. Engineers write issues the way they already do, optionally adding acceptance criteria, file hints, or test expectations to guide the agent. Orbi posts progress updates back on the issue, links the resulting pull request, and summarizes what it changed and why. When the work is done, the pull request description explains the approach, lists the files touched, and reports which checks passed. This keeps context attached to the code history so future readers can understand why a change was made.

Pricing. Orbi offers a free start: the first three merged pull requests are free and no credit card is required. After that, paid plans are available for teams that want ongoing usage on Orbi Cloud, and the open-source edition can always be self-hosted under the AGPL-3.0 license. Pricing details are published at https://orbi.build/pricing/.

Community and support. Orbi is developed in the open on GitHub, where users can file issues, propose features, and follow releases at https://github.com/orbi-build/orbi/releases. There is a public Telegram community for questions and feedback, and the team can be reached at support@orbi.build. Documentation for language models and agents is also published at https://orbi.build/llms.txt so that other AI tools can understand how Orbi works.

Why Orbi. Many AI coding tools live inside an editor and help one developer at a time. Orbi takes a different approach by living inside the team's shared workflow. It treats the issue as the specification, the repository as the context, the CI pipeline as the referee, and the pull request as the deliverable. That makes its work reviewable, auditable, and easy to adopt incrementally: start with one repository and a handful of small issues, see the pull requests it produces, and expand from there. For teams that want an autonomous software engineering agent without giving up control, transparency, or the option to self-host, Orbi offers a practical, open-source path from issue to merged pull request.