mattkir/gemini-mcp-server

Gemini MCP Server

A Model Context Protocol (MCP) server that provides access to Google's Gemini AI models, including the latest Gemini 2.5 Pro Preview.

Features

• gemini_query: Send single queries to Gemini models

• gemini_chat: Have conversations with chat history

• Support for multiple Gemini models including the latest Gemini 1.5 Pro

• Configurable generation parameters (temperature, max tokens, etc.)

• Usage metrics and response metadata

Installation

1. Install dependencies:

   cd /Users/matthewkirchoff/gemini-mcp-server
   npm install
   

2. Set up your API key: Your API key is already configured in the .env file, but you can update it if needed:

   echo "GEMINI_API_KEY=your_api_key_here" > .env
   

3. Test the server:

   npm start
   

Claude Desktop Configuration

To use this MCP server with Claude Desktop, add the following configuration to your Claude Desktop config file:

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

{
  "mcpServers": {
    "gemini": {
      "command": "node",
      "args": ["/Users/matthewkirchoff/gemini-mcp-server/src/index.js"],
      "env": {
        "GEMINI_API_KEY": "your_gemini_api_key_here"
      }
    }
  }
}

Note: Make sure to replace "your_gemini_api_key_here" with the actual value of your Gemini API key, which should be set in the environment. Do not hardcode your API key in the configuration file.

Available Tools

1. gemini_query

Send a single query to Gemini and get a response.

Parameters:

• prompt (required): The query to send to Gemini
• model (optional): Gemini model to use (default: "gemini-2.0-flash-exp")
• temperature (optional): Response creativity (0.0-2.0, default: 1.0)
• maxOutputTokens (optional): Max response length (1-8192, default: 2048)
• topP (optional): Nucleus sampling (0.0-1.0, default: 0.95)
• topK (optional): Top-k sampling (1-40, default: 40)

2. gemini_chat

Have a conversation with Gemini using chat history.

Parameters:

• messages (required): Array of conversation messages
• model (optional): Gemini model to use (default: "gemini-2.0-flash-exp")
• temperature (optional): Response creativity (0.0-2.0, default: 1.0)

Message format:

{
  "role": "user|model",
  "parts": [{"text": "message content"}]
}

Supported Models

• gemini-1.5-pro-latest - Latest Pro model (recommended)
• gemini-1.5-pro - Standard Pro model
• gemini-1.5-flash-latest - Latest Flash model
• gemini-1.5-flash - Fast, efficient model
• gemini-1.5-flash-8b - Lightweight version
• gemini-pro - Basic model
• gemini-pro-vision - Vision-capable model

Example Usage

Once configured in Claude Desktop, you can use the tools like this:

Please use the gemini_query tool to ask Gemini: "What are the latest developments in quantum computing?"

or

Use gemini_chat to have a conversation about machine learning, starting with asking about neural networks.

Troubleshooting

1. API Key Issues: Make sure your GEMINI_API_KEY is valid and has the necessary permissions
2. Model Not Found: Check if the model name is correct - use one of the supported models listed above
3. Rate Limits: Gemini API has rate limits; wait a moment before retrying
4. Large Responses: If responses are cut off, increase maxOutputTokens
5. Model Availability: Some newer experimental models may not be available in all regions

Development

To run in development mode with auto-restart:

npm run dev

Security Notes

• Keep your API key secure and never share it
• The API key is stored in the .env file and passed through environment variables
• Consider using API key rotation for production use

License

MIT License