Gemini MCP Server

What is Gemini MCP Server?

Gemini MCP Server is a Model Control Protocol (MCP) server that integrates with Google's Gemini API. Written in Go and compiled to a single self-contained binary, it provides tools for code analysis, general queries, and search with grounding. It is designed for developers using MCP-compatible clients like Claude Desktop.

How to use Gemini MCP Server?

Clone the repository and build the binary with go build -o mcp-gemini. Set the GEMINI_API_KEY environment variable and optionally specify a model. Configure the server in your MCP client (e.g., Claude Desktop) by adding a JSON entry with the binary path and environment variables (all config must be in the env section of the client configuration, not as system environment variables).

Key features of Gemini MCP Server

• Single self-contained binary with no dependencies
• Dynamic model access: fetches latest Gemini models at startup
• Advanced caching system with configurable TTL
• Seamless file integration with intelligent MIME detection
• Robust error handling, automatic retries, and graceful degradation
• Full support for code analysis, general queries, and search with grounding

Use cases of Gemini MCP Server

• Code analysis and review with file attachments and caching
• Creative writing with custom system prompts
• Factual research using Google Search integration
• Complex reasoning tasks with thinking mode enabled
• Multi-file project analysis with cost-efficient caching

FAQ from Gemini MCP Server

What are the prerequisites to use Gemini MCP Server?

You need a Google Gemini API key and Go installed to build the binary. No additional dependencies are required.

How do I configure Gemini MCP Server for Claude Desktop?

Add a JSON entry to your Claude Desktop configuration file (on macOS: ~/Library/Application Support/Claude/claude_desktop_config.json; on Windows: %APPDATA%\Claude\claude_desktop_config.json) with the binary path and all environment variables (e.g., GEMINI_API_KEY, GEMINI_MODEL) inside the env block. System environment variables are not passed to the client.

What tools does Gemini MCP Server provide?

It provides three tools: gemini_ask for code analysis and general queries with optional file context, gemini_search for grounded answers with Google Search, and gemini_models to list all available Gemini models with capabilities and caching support.

How does caching work in Gemini MCP Server?

Caching is enabled via the use_cache parameter and a configurable cache_ttl (e.g., "30m"). Files are processed once and cached, making subsequent queries faster and reducing API costs by an estimated 40–60% for extended reviews.

Which Gemini models are supported?

Models are dynamically fetched from the Google API at startup. Recommended models include gemini-2.5-pro-exp-03-25 for reasoning, gemini-2.0-flash-001 for cached tasks, and gemini-2.5-flash-preview-04-17 for search.