MCP Gemini Review

Note: This is a fork of mcp-gemini-cli by choplin, enhanced with code review capabilities.

A MCP server wrapper for Google's Gemini CLI that enables AI assistants to use Gemini's capabilities through the Model Context Protocol, with added code review functionality.

What it does

This server exposes four tools that interact with Gemini CLI:

• googleSearch: Asks Gemini to perform a Google search using your query
• chat: Sends prompts directly to Gemini for general conversations
• analyzeFile: Analyzes files (images, PDFs, text) using Gemini's multimodal capabilities
• codeReview: Reviews code files using customizable criteria from GEMINI_REVIEW.md

Prerequisites

• Gemini CLI installed and configured (optional with --allow-npx flag)

🚀 Quick Start with Claude Code

1. Add the MCP server

claude mcp add -s project gemini-review -- npx mcp-gemini-review --allow-npx

Or configure your MCP client with the settings shown in the Installation Options section below.

2. Try it out

Example prompts:

• Search: "Search for the latest TypeScript 5.0 features using Google"
• Chat: "Ask Gemini to explain the difference between async/await and promises in JavaScript"
• File Analysis: "Ask Gemini to analyze the image at /path/to/screenshot.png"
• Code Review: "Review the code in src/main.js"

🔧 Installation Options

Using npx with --allow-npx flag

{
  "mcpServers": {
    "mcp-gemini-review": {
      "command": "npx",
      "args": ["mcp-gemini-review", "--allow-npx"]
    }
  }
}

Local Development

1. Clone and install:

git clone https://github.com/cotrin8672/mcp-gemini-review
cd mcp-gemini-review
bun install

1. Add to Claude Desktop config:

{
  "mcpServers": {
    "mcp-gemini-review": {
      "command": "bun",
      "args": ["run", "/path/to/mcp-gemini-review/index.ts"]
    }
  }
}

🛠️ Available Tools

1. googleSearch

Performs a Google search using Gemini CLI.

Parameters:

• query (required): The search query
• limit (optional): Maximum number of results
• sandbox (optional): Run in sandbox mode
• yolo (optional): Skip confirmations
• model (optional): Gemini model to use (default: "gemini-2.5-pro")

2. chat

Have a conversation with Gemini.

Parameters:

• prompt (required): The conversation prompt
• sandbox (optional): Run in sandbox mode
• yolo (optional): Skip confirmations
• model (optional): Gemini model to use (default: "gemini-2.5-pro")

3. analyzeFile

Analyze files using Gemini's multimodal capabilities.

Supported file types:

• Images: PNG, JPG, JPEG, GIF, WEBP, SVG, BMP
• Text: TXT, MD, TEXT
• Documents: PDF

Parameters:

• filePath (required): The absolute path to the file to analyze
• prompt (optional): Additional instructions for analyzing the file
• sandbox (optional): Run in sandbox mode
• yolo (optional): Skip confirmations
• model (optional): Gemini model to use (default: "gemini-2.5-pro")

4. codeReview

Review code files using customizable criteria.

Parameters:

• filePath (required): The absolute path to the code file to review

Custom Review Criteria:

Create a GEMINI_REVIEW.md file in your project directory to define custom review criteria. The tool will search for this file in the same directory as the code being reviewed and its parent directories.

Example GEMINI_REVIEW.md:

# Code Review Criteria

Please review the code for:
1. Adherence to our team's naming conventions
2. Proper error handling patterns
3. Security best practices
4. Performance optimizations

💡 Example Prompts

Try these prompts to see mcp-gemini-review in action:

• Search: "Search for the latest TypeScript 5.0 features using Google"
• Chat: "Ask Gemini to explain the difference between async/await and promises in JavaScript"
• File Analysis: "Ask Gemini to describe what's in this image: /Users/me/Desktop/screenshot.png"
• Code Review: "Review the code in src/main.js"

🛠️ Example Usage

googleSearch

// Simple search
googleSearch({ query: "latest AI news" });

// Search with limit
googleSearch({
  query: "TypeScript best practices",
  limit: 5,
});

chat

// Simple chat
chat({ prompt: "Explain quantum computing in simple terms" });

// Using a different model
chat({
  prompt: "Write a haiku about programming",
  model: "gemini-2.5-flash",
});

analyzeFile

// Analyze an image
analyzeFile({ 
  filePath: "/path/to/image.png",
  prompt: "What objects are in this image?"
});

// Analyze a PDF
analyzeFile({
  filePath: "/path/to/document.pdf",
  prompt: "Summarize the key points in this document"
});

// General analysis without specific instructions
analyzeFile({ filePath: "/path/to/file.jpg" });

codeReview

// Review a JavaScript file
codeReview({ filePath: "/path/to/src/main.js" });

// Review a Python file
codeReview({ filePath: "/path/to/app.py" });

// The tool will automatically look for GEMINI_REVIEW.md 
// in the project directory for custom review criteria

📝 Development

Note: Development requires Bun runtime.

Run in Development Mode

bun run dev

Run Tests

bun test

Build for Production

# Development build
bun run build

# Production build (minified)
bun run build:prod

Linting & Formatting

# Lint code
bun run lint

# Format code
bun run format

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

📋 Changelog

[0.3.1] - 2025-07-03

Fixed

• Fixed Windows compatibility issue with which command

[0.3.0] - 2025-07-02

Breaking Changes

• Tool names: geminiChat → chat, geminiAnalyzeFile → analyzeFile
• Package name: @choplin/mcp-gemini-cli → mcp-gemini-cli

New Features

• analyzeFile tool for images (PNG/JPG/GIF/etc), PDFs, and text files

[0.2.0] - Previous

• Initial release with googleSearch and geminiChat tools

🔗 Related Links

• Model Context Protocol (MCP)
• Gemini CLI
• Bun Runtime