You ask for search, reasoning, processing or file analysis, and the server picks the Gemini variant to run it on — which matters mostly for cost, since gem_process routes to the cheapest variant and gem_reason to the one that shows its work. The uniform file_path parameter is the other reason to reach for it: the same call handles a PDF, a source file or an image.
A Gemini API server with four task-shaped tools. Each one maps to the Gemini variant suited to it — grounded search, step-by-step reasoning, or cheap bulk processing — and every tool accepts a file_path so documents, code and images go in the same way.
- gem_search — factual questions answered with Gemini's search integration, so the response is grounded rather than recalled
- gem_reason — step-by-step reasoning for maths, science and code problems, with show_steps to see the working
- gem_process — the token-cheapest path, for summarising and extracting from text or files at volume
- gem_analyze — file analysis with the model picked automatically from the file type: images, code, or plain text
- A file_path parameter on every tool, with the MIME type detected for you
- Retries with exponential backoff when the API rate-limits you
A Gemini API key from Google AI Studio, supplied as GEMINI_API_KEY in the environment or a .env file. Node.js 16 or newer. Install through the Smithery CLI, or clone, npm install and npm run build.
One command plus a key — npx -y @smithery/cli@latest install @PV-Bhat/gemsuite-mcp --client claude, then supply credentials