Gemini MCP Server

A Model Context Protocol (MCP) server that provides a bridge to Google's Gemini CLI. This allows MCP-compliant clients (like Claude Desktop) to interact with the Gemini CLI to start new chat sessions or continue existing ones directly from the client interface.

Features

• Start New Sessions: Initiate a new conversation with a specific Gemini model.
• Continue Sessions: Reply to existing sessions using their Session ID.
• Session Management: Automatically detects the latest session ID if one is not provided for replies.
• Model Selection: Supports specifying different Gemini models (e.g., gemini-3-pro).

Prerequisites

Before using this server, ensure you have the following installed:

1. Node.js: v18 or higher.
2. Gemini CLI: You must have the gemini CLI tool installed and configured on your system.
   • Ensure gemini is in your system's PATH.
   • Ensure you have authenticated or set up necessary API keys for the gemini CLI to work.

Installation

You can run this server directly using npx or by building it locally.

Local Setup

1. Clone the repository:

   git clone <repository-url>
   cd gemini-mcp-server

2. Install dependencies:

   npm install

3. Build the project:

   npm run build

Configuration

To use this with an MCP client (e.g., Claude Desktop), add the server configuration to your client's settings file.

Claude Desktop Configuration

Add the following to your claude_desktop_config.json (usually located at ~/Library/Application Support/Claude/claude_desktop_config.json on macOS or %APPDATA%\Claude\claude_desktop_config.json on Windows):

Using source (for development): MCP (Model Context Protocol) server for interacting with Google's Gemini CLI.

This server allows AI assistants like Claude to use Google's Gemini through the MCP protocol.

Prerequisites

• Node.js 18+
• Gemini CLI installed and authenticated

# Install Gemini CLI
npm install -g @anthropic-ai/gemini-cli
# or
brew install gemini-cli

# Authenticate
gemini auth

Installation

Global Installation

npm install -g @2lab.ai/gemini-mcp-server

Run directly with npx

npx @2lab.ai/gemini-mcp-server

Usage

Running the server

After global installation:

gemini-mcp-server

Or with npx:

npx @2lab.ai/gemini-mcp-server

Testing with MCP Inspector

You can test and debug the server using the MCP Inspector:

npx @modelcontextprotocol/inspector npx @2lab.ai/gemini-mcp-server

Or if globally installed:

npx @modelcontextprotocol/inspector gemini-mcp-server

Claude Desktop Configuration

Add the following to your Claude Desktop configuration file:

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

{
  "mcpServers": {
    "gemini": {
      "command": "npx",
      "args": ["-y", "@2lab.ai/gemini-mcp-server"]
    }
  }
}

Or if globally installed:

{
  "mcpServers": {
    "gemini": {
      "command": "gemini-mcp-server"
    }
  }
}

Claude Code Configuration

Add to your Claude Code settings:

{
  "mcpServers": {
    "gemini": {
      "command": "npx",
      "args": ["-y", "@2lab.ai/gemini-mcp-server"]
    }
  }
}

Available Tools

gemini

Start a new Gemini session with a prompt.

Parameters:

• prompt (required): The prompt to start the session with
• model Optional: The model to use (e.g., 'gemini-3-pro', 'gemini-3-flash').

Returns: Response text and new Session ID

gemini-reply

Continue an existing Gemini session.

Parameters:

• prompt (required): The prompt to continue the conversation
• sessionId (optional): The session ID to continue. If not provided, uses the latest session
• model (optional): The model to use for this turn

Returns: Response text

Development

# Clone the repository
git clone https://github.com/2lab-ai/gemini-mcp-server.git
cd gemini-mcp-server

# Install dependencies
npm install

# Build
npm run build

# Run locally
npm start

CI/CD

This project uses GitHub Actions to automatically publish to npm when changes are pushed to the main branch.

To enable automatic publishing:

1. Generate an npm access token from npmjs.com
2. Add the token as a secret named NPM_TOKEN in your GitHub repository settings
3. Bump the version in package.json before pushing to trigger a publish

License

ISC