Skip to content

Latest commit

 

History

624 Commits

Folders and files

Repository files navigation

npm License Price Providers

co-maintainer

co-maintainer

An open source PR reviewer that learns how your repo is reviewed, then reviews like that.

It reads your code, your merged PRs and the review comments on them, and writes review guides as plain Markdown. Every review follows those guides. It runs on your machine or on your own server, with your own model key.

Website · Docs · Newsletter

npm install -g co-maintainer

co-maintainer probe owner/repo        # reads only, spends nothing
co-maintainer init owner/repo --include-codebase --include-pull-requests \
  --max-pr-months=12 --max-commits=500
co-maintainer review owner/repo 123   # review a pull request

Tip

co-maintainer review with no arguments reviews your local branch before you open the PR. co-maintainer serve runs the GitHub App and a dashboard for every repository on your own server.

What it does

A P1 finding on PR 54 Asking the reviewer what it meant
New, still open and closed findings Review policy options
probe output before init review --json output for agents

How it works

flowchart LR
  A[Your repo<br>code, PRs, review comments] --> B[init]
  B --> C[Review guides<br>plain Markdown]
  C --> D[review]
  E[New merged PRs] --> F[sync] --> C
Loading

A low model reads the history and extracts evidence. A high model writes the guides and does the review. You pick both models. Read what it learned with co-maintainer view.

Pricing

Solo Team Company
$0 $0 per seat $0 forever
The CLI on a laptop One server, guides shared with --remote Your infrastructure, MIT licensed

You only pay your model provider. No telemetry from the CLI, serve or the dashboard. How we earn: donations and thanks.

Benchmark

core_v3: 30 PRs in one target repo, 14 with seeded defects and 16 clean controls, so precision is measured too.

Recall Precision F1 Cost / PR Time / PR
0.857 0.353 0.500 $0.0045 28.9s

deepseek-v4-flash + gpt-5.6-luna, judge gpt-5.6-luna, 2026-09-14. A small dataset, read it as a signal. Details: benchmark/core_v3.

Development

npm install
npm test                 # unit, script and benchmark tests
npm run check            # tsc --noEmit
npm run docs:build       # rebuild docs/ from docs/md/
npm run e2e              # end-to-end CLI loop
npm run review-local-e2e # fake-AI local review loop
npm run dashboard:smoke  # browser smoke test
npm run e2e:simulate     # init, sync, review and App flows in a simulated GitHub, see CONTRIBUTING.md

Made by Murat Kirazkaya. If it helps you, a star or a thank you is how this project gets paid.

Packages

Contributors

Languages