Git Reviewer

License: MIT MCP Protocol

Stop manually copying and pasting massive code diffs into your AI chat window.

Git-Reviewer is a lightweight, secure Model Context Protocol (MCP) server that connects LLM assistants (like Claude Desktop, Cursor, and Windsurf) directly to the GitHub API. It enables your AI to fetch pull request metadata and raw diff files autonomously, producing fast, context-aware, and highly accurate code reviews directly within your development workspace.

---

⚡ The Context Switching Problem (Solved)

flowchart LR
    DEV[Developer] -->|Ask for PR review| AI[AI Assistant]
    AI -->|MCP JSON-RPC| SERVER[Git-Reviewer MCP Server]

    SERVER --> META[fetch_pr_metadata]
    SERVER --> DIFF[fetch_pr_diff]

    META -->|Authorized API call| GITHUB[GitHub REST API]
    DIFF -->|Authorized API call| GITHUB

    GITHUB --> DATA[PR title, author, state, description, and raw diff]
    DATA --> SERVER
    SERVER -->|Structured review context| AI
    AI -->|Actionable review findings| DEV

    AI -. Optional: publish line comment .-> COMMENT[post_pr_review_comment]
    COMMENT -.-> GITHUB

Instead of manually navigating tabs, copying hundreds of lines of diff code, and hitting token limits, you can now simply type:

"Analyze the code changes in react/react PR #36818 and check for architectural issues."

Your assistant will query the server, read the raw diff, and deliver a structured code review instantly.

---

✨ Features

• 💨 High-Fidelity Diffs (fetch_pr_diff): Fetches raw, line-by-line .diff code changes, optimized for LLM comprehension.
• 📋 Rich Context (fetch_pr_metadata): Retrieves PR Title, Author, Description, and State to give the AI crucial background.
• ✍️ Active Review Comments (post_pr_review_comment): Allows the AI to write line-level feedback directly onto the PR from the chat interface. (Automatically resolves head commit SHA if omitted).
• 🛡️ Secure Token Storage: Uses standard local .env configuration. Your GitHub Personal Access Token is never committed or shared.

---

🚀 Installation & Local Setup

1. Clone & Initialize Environment

Set up a clean Python environment:

# Clone the repository
git clone https://github.com/Suchit-007/git-reviewer.git
cd git-reviewer

# Create and activate virtual environment
python -m venv .venv
# On Windows:
.venv\Scripts\activate
# On macOS/Linux:
source .venv/bin/activate

# Install required packages
pip install -r requirements.txt

2. Configure Environment Variables

Create a .env file in the root folder (or copy .env.example):

GITHUB_TOKEN=your_personal_access_token_here

Note: Using a token avoids GitHub API rate limiting on public repositories and enables reading from private repositories.

---

⚙️ IDE Integration

Claude Desktop

Add the configuration to your claude_desktop_config.json file:

• Windows: %APPDATA%\Claude\claude_desktop_config.json
• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "git-reviewer": {
      "command": "python",
      "args": [
        "C:\\absolute\\path\\to\\git-reviewer\\src\\server.py"
      ],
      "env": {
        "GITHUB_TOKEN": "your_github_token_here"
      }
    }
  }
}

Cursor

1. Go to Settings -> Features -> MCP.
2. Click + Add New MCP Server.
3. Configure:
   • Name: git-reviewer
   • Type: command
   • Command: python C:\absolute\path\to\git-reviewer\src\server.py
4. Click Save.

---

🧪 Development & Testing

You can test the server locally using the MCP Inspector tool:

# Activate virtual environment
.venv\Scripts\activate

# Run the inspector
fastmcp dev inspector src/server.py

This opens a local developer interface at http://localhost:6274 to test the tools interactively.

---

📄 License

This project is open-source and licensed under the MIT License.