A lightweight MCP server that enables programming assistants to interact with Google Gemini AI through the official CLI, providing zero API costs and seamless integration with Claude Code, Cursor, VS Code, and other MCP-compatible clients.

Installation

Installation via PyPI

pip install gemini-bridge
claude mcp add gemini-bridge -s user -- uvx gemini-bridge

From Source Code

git clone https://github.com/shelakh/gemini-bridge.git
cd gemini-bridge
uvx --from build pyproject-build
pip install dist/*.whl
claude mcp add gemini-bridge -s user -- uvx gemini-bridge

Development Installation

git clone https://github.com/shelakh/gemini-bridge.git
cd gemini-bridge
pip install -e .
claude mcp add gemini-bridge-dev -s user -- python -m src

Configuration

Cursor

{
  "mcpServers": {
    "gemini-bridge": {
      "command": "uvx",
      "args": ["gemini-bridge"],
      "env": {}
    }
  }
}

VS Code

{
  "servers": {
    "gemini-bridge": {
      "type": "stdio",
      "command": "uvx",
      "args": ["gemini-bridge"]
    }
  }
}

With Custom Timeout

{
  "mcpServers": {
    "gemini-bridge": {
      "command": "uvx",
      "args": ["gemini-bridge"],
      "env": {
        "GEMINI_BRIDGE_TIMEOUT": "120"
      }
    }
  }
}

Available Tools

Features

• Direct integration with Gemini CLI at zero API cost
• Simple MCP tools for basic queries and file analysis
• Stateless operation with no sessions or caching
• Production-ready with robust error handling and 60-second timeouts
• Minimal dependencies requiring only mcp>=1.0.0 and Gemini CLI
• Universal MCP compatibility for Claude Code, Cursor, VS Code, and other clients
• Installation support via both uvx and traditional pip

Environment Variables

Optional

• GEMINI_BRIDGE_TIMEOUT - Set custom timeout for CLI operations (default: 60 seconds)
• GEMINI_BRIDGE_MAX_INLINE_TOTAL_BYTES - Maximum bytes for inline file transfer

Usage Examples

What authentication patterns are used in this codebase?

Review these auth files for security issues

Find authentication patterns in this codebase

Analyze these auth files and suggest improvements

Compare these database implementations and recommend the best approach

Notes

Requires installation and authentication of Google Gemini CLI (npm install -g @google/gemini-cli && gemini auth login). Supports model selection between 'flash' and 'pro', with inline mode for small files and at_command mode for operations with larger files. Includes file size protection with limits of ~256 KB per file and ~512 KB per request.