Code Reviewer MCP Server

A GitHub Pull Request Code Review server using Model Context Protocol

This MCP server provides automated code review capabilities for GitHub pull requests by integrating with OpenRouter's AI models.

Features

• Automated code reviews for GitHub pull requests
• Detailed feedback on code quality, style and potential issues
• Integration with OpenRouter's AI models (currently using Gemini 2.5 Pro)
• Comprehensive logging for debugging

Requirements

• A GitHub Personal Access Token (PAT) with repo scope for private repositories or public_repo scope for public repositories. This PAT is required to fetch pull request data.
• An OpenRouter API Key.

Current Limitations

⚠️ Important: Currently this server only works with:

• Open pull requests (not draft PRs)

Installation

1. Clone this repository:

git clone https://github.com/joaomj/code-reviewer-mcp.git
cd code-reviewer-mcp

2. Install dependencies:

npm install

3. Set required environment variables:

export GITHUB_PAT=your_github_personal_access_token
export OPENROUTER_API_KEY=your_openrouter_api_key

4. Build the server:

npm run build

Usage

To start the MCP server:

node build/index.js

The server will be available to any MCP-compatible client (like Claude Desktop).

Configuration

Add to your MCP client configuration:

{
  "mcpServers": {
    "code-reviewer-mcp": {
      "command": "node",
      "args": ["/path/to/code-reviewer-mcp/build/index.js"],
      "env": {
        "GITHUB_PAT": "your_github_token",
        "OPENROUTER_API_KEY": "your_openrouter_key"
      }
    }
  }
}

Development

For development with auto-rebuild:

npm run watch

To run tests:

npm test

Debugging

Use the built-in logging which outputs to stderr with detailed information about:

• GitHub API requests
• OpenRouter API calls
• Processing steps
• Any errors encountered

For more advanced debugging, use the MCP Inspector:

npm run inspector

Next Steps

Planned improvements and features:

• Draft pull request analysis
• Customizable review templates
• Support for multiple AI models (OpenAI, Claude, etc.)
• File-specific feedback
• Rate limiting and caching

Contributing

Suggestions are welcome. Please open an issue first to discuss what you would like to change.