Gemini Interactions MCP

A FastMCP server for stateful Gemini conversations with automatic Google Search grounding.

Features

• Stateful Conversations: Maintain context across queries via interaction_id
• Google Search Grounding: Model automatically searches the web when needed
• URL Context: Parse and analyze linked web pages
• Thinking Levels: Control reasoning depth (minimal, medium, high)

Setup

uv sync
cp .env.example .env
# Edit .env with your GEMINI_API_KEY

Get your API key from: https://aistudio.google.com/app/apikey

Usage

uv run python server.py

Tools

Tool	Thinking Level	Description
search	minimal	Quick web search, structured results
ask	medium	Balanced grounded answers
ask_thinking	high	Deep reasoning with grounding

All tools support interaction_id for stateful follow-up conversations.

Claude Desktop Integration

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "gemini": {
      "command": "uv",
      "args": ["run", "python", "/path/to/gemini-interactions-mcp/server.py"],
      "env": {
        "GEMINI_API_KEY": "your_api_key_here"
      }
    }
  }
}

License

MIT