mr-review

AI-powered merge request review tool. Runs locally via Docker, connects to GitLab, GitHub, Gitea, Forgejo, or Bitbucket, and works with Claude, OpenAI, or any OpenAI-compatible model.

Full documentation → bedrock-python.github.io/mr-review

Tip

Deploying this with an AI assistant? Hand it one page instead of the whole site: the smallest compose file that runs, every environment variable and its real default, the rules that break a deployment when they are broken, the mistakes models make in a compose file, and a map of which page to fetch for the rest. Every docs page is also served as raw Markdown at its own URL, and a Copy page button at the top of each one hands it straight to a chat window.

---

Quick start

mkdir mr-review && cd mr-review
curl -O https://raw.githubusercontent.com/bedrock-python/mr-review/master/deploy/all-in-one/docker-compose.yml
docker compose up -d

Open http://localhost:17240, add an AI provider and a VCS host, then pick an MR to review.

That's it — no accounts, no cloud, no data leaves your machine.

---

How it works

Pick a merge request and let the AI walk through it in four stages:

Stage	What happens
Brief	Choose a review preset (thorough / security / style / performance) and add custom instructions
Dispatch	AI reviews the diff — comments appear as they stream in
Polish	Edit, keep, or dismiss individual comments before posting
Post	Approved comments are sent back to the MR as inline review notes

---

Deployment

All-in-one (recommended)

Single container, single port. Easiest way to get started.

mkdir mr-review && cd mr-review
curl -O https://raw.githubusercontent.com/bedrock-python/mr-review/master/deploy/all-in-one/docker-compose.yml
docker compose up -d
# → http://localhost:17240

Standard (separate services)

API and UI run as separate containers — useful if you want more control over networking or scaling.

mkdir mr-review && cd mr-review
curl -O https://raw.githubusercontent.com/bedrock-python/mr-review/master/deploy/standard/docker-compose.yml
docker compose up -d
# API → http://localhost:17241
# UI  → http://localhost:17242

Docker images are published to GitHub Container Registry:

ghcr.io/bedrock-python/mr-review/all-in-one:latest
ghcr.io/bedrock-python/mr-review/api:latest
ghcr.io/bedrock-python/mr-review/web-app:latest

---

Configuration

All configuration is done through the UI after first launch:

1. Add an AI provider — Claude (Anthropic), OpenAI, or any OpenAI-compatible endpoint
2. Add a VCS host — GitLab, GitHub, Gitea, Forgejo, or Bitbucket with a personal access token
3. Pick a repository and MR — start reviewing

See the configuration guide for environment variables and advanced options.

---

Documentation

For AI agents	The whole tool on one page, written for a coding assistant
Installation	Docker options, ports, data persistence
Quick start	Get up and running in 2 minutes
Configuration	AI providers, VCS hosts, environment variables
Contributing	Running the services locally, checks, releases

---

Development

# Start backend (:8000) + frontend dev server (:5173)
make dev

# Run tests
make run-tests

# Format all services
make fmt-services

# See all available commands
make help

Service READMEs:

• services/mr-review/README.md — Python FastAPI backend
• services/web-app/README.md — React TypeScript frontend