mcp -gemini-pro

This MCP server bundles Gemini models for integration with MCP-compatible clients like Claude Desktop. It exposes a set of capabilities including advanced thinking models, real-time grounding, JSON-mode outputs, and multimodal vision. To use it, configure Claude Desktop (or any MCP client) to point at the local node-based server by providing the command and arguments above, along with your Gemini API key. Once configured, you can issue natural language prompts to request text generation, image analysis, or structured outputs, and leverage features like JSON-mode for strict schema responses, or grounding for up-to-date information from the web.

Available tools include: generate_text for advanced thinking with step-by-step reasoning, analyze_image for multimodal vision tasks, count_tokens to estimate costs, list_models to view Gemini model availability, embed_text for vector representations, and get_help for usage guidance. These tools work through the MCP protocol, enabling seamless interactions with Gemini models, memory, and grounding features within your existing MCP workflow.

Prerequisites:\n- Node.js 16+ installed on your system (https://nodejs.org/)\n- Git installed (https://git-scm.com/)\n- A Gemini API key (Google AI Studio)\n\nStep-by-step:\n1) Clone the repository:\n git clone https://github.com/your-org/mcp-server-gemini-pro.git\n cd mcp-server-gemini-pro\n\n2) Install dependencies and build:\n npm install\n npm run build\n\n3) Run the server (local development):\n node dist/enhanced-stdio-server.js\n\n4) Optional: install globally (recommended for CLI usage):\n npm install -g mcp-server-gemini-pro\n # then run the server as: mcp-server-gemini-pro\n\n5) Configure your environment:\n - Set GEMINI_API_KEY in your environment or via a .env file.\n - Ensure the path to dist/enhanced-stdio-server.js is correct in your MCP client configuration.\n\n6) Connect via Claude Desktop or another MCP client using the provided mcpServers gemini entry.

Environment variables: GEMINI_API_KEY is required. Other optional settings include LOG_LEVEL, RATE_LIMIT_* values, and REQUEST_TIMEOUT. If you encounter issues connecting from Claude Desktop, ensure the server path in your MCP client config matches the built dist/enhanced-stdio-server.js file. When running locally, you may need to provide the absolute path to the server file in your MCP client configuration. For production deployments, consider running with a process manager (e.g., pm2) and setting NODE_ENV=production. If you update models or API keys, restart the MCP server to apply changes.