code-review-mcp
Enables AI assistants to review GitHub and GitLab pull requests and merge requests, supporting multiple transports and custom review rules.
README
Code Review MCP Server
English | 中文
MCP (Model Context Protocol) server for code review. Enables AI assistants to review GitHub/GitLab Pull Requests and Merge Requests.
✨ Features
- 🔍 Multi-platform: Supports both GitHub and GitLab (including self-hosted)
- 🚀 Multiple Transports: Supports stdio, SSE, and WebSocket protocols
- 📦 Easy Install: Quick install via
uvxorpip - 🐳 Containerized: Docker image available
- ☁️ Cloud Deploy: One-click Smithery deployment
- 🔒 Security First: Environment variable configuration, no data persistence
🚀 Quick Start
Option 1: Using uvx (Recommended)
# Run directly, no installation needed
uvx code-review-mcp
Option 2: Using pip
pip install code-review-mcp
# Run the server
code-review-mcp
# (Optional) Install Cursor rules to your project
code-review-mcp init-rules
Option 3: From Source
git clone https://github.com/OldJii/code-review-mcp.git
cd code-review-mcp
pip install -e .
code-review-mcp
🔧 Configuration
Environment Variables
| Variable | Description | Required |
|---|---|---|
GITHUB_TOKEN |
GitHub personal access token | When using GitHub |
GITLAB_TOKEN |
GitLab personal access token | When using GitLab |
GITLAB_HOST |
GitLab host URL | For self-hosted (default: gitlab.com) |
Getting Tokens
GitHub
# Option 1: Using gh CLI (Recommended)
brew install gh
gh auth login
# Option 2: Manual Token Creation
# Visit https://github.com/settings/tokens
# Create Personal Access Token with 'repo' scope
export GITHUB_TOKEN="your-token-here"
GitLab
# Option 1: Using glab CLI (Recommended)
brew install glab
glab auth login
# For self-hosted GitLab
glab auth login --hostname gitlab.yourcompany.com
# Option 2: Manual Token Creation
# Visit GitLab -> Settings -> Access Tokens
# Create token with 'api' scope
export GITLAB_TOKEN="your-token-here"
export GITLAB_HOST="gitlab.yourcompany.com" # For self-hosted
📱 Client Configuration
Cursor
Edit ~/.cursor/mcp.json:
{
"mcpServers": {
"code-review": {
"command": "uvx",
"args": ["code-review-mcp"],
"env": {
"GITHUB_TOKEN": "your-github-token",
"GITLAB_TOKEN": "your-gitlab-token"
}
}
}
}
Claude Desktop
Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):
{
"mcpServers": {
"code-review": {
"command": "uvx",
"args": ["code-review-mcp"],
"env": {
"GITHUB_TOKEN": "your-github-token",
"GITLAB_TOKEN": "your-gitlab-token"
}
}
}
}
SSE Mode (Remote Deployment)
# Start SSE server
code-review-mcp --transport sse --port 8000
Client configuration:
{
"mcpServers": {
"code-review": {
"url": "http://your-server:8000/sse"
}
}
}
WebSocket Mode (Remote Deployment)
# Start WebSocket server
code-review-mcp --transport websocket --port 8000
Client configuration:
{
"mcpServers": {
"code-review": {
"url": "ws://your-server:8000/ws"
}
}
}
🐳 Docker Deployment
Build Image
docker build -t code-review-mcp .
Run Container
stdio mode
docker run -i --rm \
-e GITHUB_TOKEN="your-token" \
code-review-mcp
SSE mode
docker run -d --rm \
-e GITHUB_TOKEN="your-token" \
-p 8000:8000 \
code-review-mcp --transport sse
🔨 MCP Tools
Rules
| Tool | Description |
|---|---|
get_review_rules |
Get review rules (builtin + custom project rules) |
Information Retrieval
| Tool | Description |
|---|---|
get_pr_info |
Get PR/MR details (title, description, branches) |
get_pr_changes |
Get code changes (diff), supports file type filtering |
extract_related_prs |
Extract related PR/MR links from description |
Adding Comments
| Tool | Description |
|---|---|
add_inline_comment |
Add inline comment to specific code line |
add_pr_comment |
Add general comment |
batch_add_comments |
Batch add comments (inline + general) |
💬 Usage Examples
Chat with Cursor or Claude:
Review GitHub PR
Review https://github.com/facebook/react/pull/12345
Review GitLab MR
Review https://gitlab.com/group/project/-/merge_requests/678
Review Self-hosted GitLab MR
Review https://gitlab.yourcompany.com/team/project/-/merge_requests/90
Review Only Specific File Types
Review this PR, only check .py and .js files:
https://github.com/owner/repo/pull/123
🧪 Debugging & Testing
Using MCP Inspector
# Run with MCP Inspector
npx @modelcontextprotocol/inspector uvx code-review-mcp
This launches a web interface where you can:
- View all available tools
- Manually call tools and inspect results
- Debug parameters and responses
Local Development
# Clone repository
git clone https://github.com/OldJii/code-review-mcp.git
cd code-review-mcp
# Install dev dependencies
pip install -e ".[dev]"
# Run tests
pytest
# Debug with Inspector
npx @modelcontextprotocol/inspector python -m code_review_mcp.server
📁 Project Structure
code-review-mcp/
├── src/
│ └── code_review_mcp/
│ ├── __init__.py # Package entry
│ ├── cli.py # CLI commands (init-rules, etc.)
│ ├── server.py # MCP server main logic
│ ├── providers.py # GitHub/GitLab providers
│ └── rules/ # Bundled Cursor rules
│ ├── code-review.mdc
│ └── code-review-en.mdc
├── pyproject.toml # Project config & PyPI publishing
├── Dockerfile # Docker build file
├── smithery.yaml # Smithery deployment config
├── CHANGELOG.md # Changelog
├── CONTRIBUTING.md # Contributing guide
└── README.md # Documentation
🎯 Cursor Rules (Recommended)
This package includes built-in code review rules for Cursor IDE. Install them to your project with one command:
# Install rules to current project
code-review-mcp init-rules
# Install to a specific directory
code-review-mcp init-rules --target /path/to/project
# Overwrite existing rules
code-review-mcp init-rules --force
# List available rules
code-review-mcp list-rules
After installation, the rules will be available in your project's .cursor/rules/ directory:
code-review.mdc- Chinese versioncode-review-en.mdc- English version
Custom Project Rules
You can define project-specific review rules that the MCP server loads at runtime. This allows each project to enforce its own coding standards during reviews.
Quick Setup:
# Generate a custom rules template
code-review-mcp init-rules --custom
This creates .code-review-rules/project-rules.md in your project. Edit it with your project-specific conventions, then configure the MCP server to load it:
{
"mcpServers": {
"code-review": {
"command": "uvx",
"args": ["code-review-mcp"],
"env": {
"GITHUB_TOKEN": "your-token",
"CODE_REVIEW_RULES_DIR": "/absolute/path/to/project/.code-review-rules"
}
}
}
}
How It Works:
- Set
CODE_REVIEW_RULES_DIRto a directory containing.mdor.mdcfiles - Or simply place a
.code-review-rules/directory in your project root (auto-discovered) - The
get_review_rulestool returns both builtin and custom rules - AI assistants use these rules when performing reviews
- Custom rules supplement (not replace) the builtin review guidelines
Environment Variable:
| Variable | Description | Required |
|---|---|---|
CODE_REVIEW_RULES_DIR |
Path to custom rules directory | No (optional) |
🤝 Contributing
Contributions welcome! See CONTRIBUTING.md for details.
📄 License
🔗 Related Links
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。