Gemini MCP Tool

The Gemini MCP Tool is an MCP server that connects an AI assistant, like Claude, with the Google Gemini CLI.

It lets you use Gemini’s powerful, large-token window for analyzing big files and codebases directly within your preferred MCP clients.

Features

• 🔗 Direct Gemini CLI Integration – Execute Gemini commands directly from Claude through MCP
• 📁 File Reference Support – Use @ syntax to analyze specific files or entire directories
• 🛡️ Sandbox Mode – Safe code execution and testing in isolated environments
• 🎯 Model Selection – Choose from different Gemini models including gemini-2.5-flash
• 💬 Natural Language Interface – Ask questions in plain English without complex command syntax
• ⚡ NPX Support – Run without installation using npx for immediate setup
• 🔧 Slash Commands – Built-in commands for Claude Code interface integration
• 📊 Large File Analysis – Leverage Gemini’s extensive context window for massive codebases

Use Cases

• Legacy Code Analysis: Analyze large, undocumented codebases where token limits prevent comprehensive understanding in a single Claude session
• Project Documentation: Generate detailed documentation for complex projects by having Gemini analyze entire directory structures and dependencies
• Code Review at Scale: Review pull requests or commits across multiple files simultaneously without hitting context limitations
• Safe Script Testing: Use sandbox mode to test potentially risky scripts or commands before implementing them in production environments

How To Use It

1. Add the MCP Server to your MCP clients like Claude:

claude mcp add gemini-cli -- npx -y gemini-mcp-tool

2. Type /mcp in the Claude Code interface to confirm that the gemini-cli MCP is active.

3. If you prefer to install the package globally, run:

npm install -g gemini-mcp-tool

Then, configure your client (like Claude Desktop) by adding the following to your mcpServers JSON object:

"gemini-cli": {
  "command": "gemini-mcp"
}

4. You can then interact with the MCP server using natural language or specific slash commands.

• ask gemini to analyze @src/main.js and explain what it does
• use gemini to summarize @. the current directory
• use gemini sandbox to install numpy and create a data visualization
• /analyze prompt:<your_prompt>: Analyzes files/directories or asks a general question. Example: /analyze prompt:@src/ summarize this directory.
• /sandbox prompt:<your_prompt>: Tests code or scripts in the sandbox. Example: /sandbox prompt:@script.py Test this script safely.
• /help: Displays help information.
• /ping: Tests the connection to the server.

FAQs

Q: Do I need to install Gemini CLI separately?
A: Yes, the Gemini CLI must be installed and configured independently. This tool acts as an MCP interface to your existing Gemini CLI installation.

Q: What’s the difference between sandbox mode and regular mode?
A: Sandbox mode provides an isolated environment for safe code execution and testing. Use it when running potentially risky scripts or commands that you want to test before implementing.

Q: How do I reference multiple files in a single request?
A: Use multiple @ references in your prompt: analyze @src/main.js @src/utils.js @package.json and explain the project structure