Open-source AI code review
Reviews that read
the whole story.
OwnRabbit is a pull request reviewer you can run yourself. A team of specialist agents reads your diff, digs into the codebase, argues it out, and leaves only the comments worth reading.
14 export async function getUser(req) {- const id = Number(req.params.id);15+ const id = req.params.id;16+ return db.query(`SELECT * FROM users WHERE id = ${id}`);17 }● Critical · security
SQL injection through the user id
id is no longer coerced to a number and is interpolated straight into SQL. Use a bound parameter.
return db.query('SELECT * FROM users WHERE id = $1', [id]);Four stages. One quiet review.
- 01
Triage
A fast model maps the PR, ranks files by risk, and skips the noise: lockfiles, generated code, assets.
- 02
Specialists
Security, correctness, performance, maintainability, and test reviewers run in parallel, each with its own brief.
- 03
Investigate
On deeper settings, reviewers read callers, types, and configs in a checkout of your branch before they speak.
- 04
Judge
A final pass throws out speculative, duplicate, and wrong findings. What survives is what a senior engineer would post.
Every repo,
your rules.
Each repository gets its own settings in the dashboard. Loud on the payments service, quiet on the docs site.
- Trigger
- every push, on @mention, or both
- Depth
- quick · standard · deep
- Reviewers
- pick which specialists run
- Budget
- cap comments, set a severity floor
- Scope
- include / exclude globs
- Rules
- your house rules, in plain English
Open source
Read the prompts. Change them.
Every prompt, rule, and pipeline stage is in the repository. Run it on your own cloud, point it at your own model, and send a PR when you make it better.