taehun-kmu/gemini-mcp-server

Gemini MCP Server

MCP (Model Context Protocol) Server with Google Gemini CLI integration for AI second opinions and validation in development workflows.

Features

• 🤖 AI Second Opinions: Get alternative perspectives from Google's Gemini AI
• 🔍 Automatic Uncertainty Detection: Triggers consultations when uncertainty is detected
• ⚡ Rate Limiting: Built-in rate limiting to prevent API abuse
• 🛡️ Configurable Models: Support for different Gemini models
• 📊 Consultation Logging: Track all AI consultations with timestamps

Quick Start

Using uvx (Recommended)

Run directly without installation:

# Run in current directory
uvx gemini-mcp-server

# Run with specific project root
uvx gemini-mcp-server --project-root /path/to/project

Local Development

# Clone the repository
git clone https://github.com/taehun-kmu/gemini-mcp-server
cd gemini-mcp-server

# Install with uv
uv pip install -e .

# Run the server
gemini-mcp-server

Prerequisites

1. Node.js 18+ - Required for Gemini CLI
2. Python 3.8+ - Required for MCP server
3. Gemini CLI - Install and authenticate:

# Install Gemini CLI
npm install -g @google/gemini-cli

# Authenticate (run once)
gemini

Configuration

Create a gemini-config.json file in your project root:

{
  "enabled": true,
  "auto_consult": true,
  "cli_command": "gemini",
  "timeout": 60,
  "rate_limit_delay": 2.0,
  "model": "gemini-2.5-flash",
  "log_consultations": true
}

Environment Variables

Override configuration with environment variables:

• GEMINI_ENABLED: Enable/disable integration
• GEMINI_AUTO_CONSULT: Enable/disable automatic consultation
• GEMINI_CLI_COMMAND: CLI command (default: "gemini")
• GEMINI_TIMEOUT: Command timeout in seconds
• GEMINI_RATE_LIMIT: Delay between consultations
• GEMINI_MODEL: Model to use (default: "gemini-2.5-flash")

MCP Tools

The server exposes three MCP tools:

1. consult_gemini: Get second opinions from Gemini

   • query: The question or topic
   • context: Additional context
   • comparison_mode: Request structured comparison format

2. gemini_status: Check integration status and statistics

3. toggle_gemini_auto_consult: Enable/disable automatic consultation

   • enable: true/false or omit to toggle

Claude Code Integration

Configure Claude Code to use this MCP server by adding to your MCP configuration:

{
  "mcpServers": {
    "gemini": {
      "command": "uvx",
      "args": ["gemini-mcp-server"],
      "env": {
        "GEMINI_ENABLED": "true"
      }
    }
  }
}

Note: Add "--project-root", "/path/to/project" to args if you need to specify a different project directory.

Development

Project Structure

gemini-mcp-server/
├── gemini_mcp/
│   ├── __init__.py
│   ├── __main__.py         # CLI entry point
│   ├── gemini_integration.py  # Gemini integration logic
│   └── server.py           # MCP server implementation
├── pyproject.toml          # Package configuration
├── README.md              # This file
├── LICENSE                # MIT License
├── setup-gemini-integration.sh  # Setup script
└── gemini-config.json     # Example configuration

Running Tests

# Install dev dependencies
uv pip install -e ".[dev]"

# Run tests
pytest

# Format code
ruff format gemini_mcp
ruff check gemini_mcp

License

MIT License - see LICENSE file for details.