Summary
Signing up for a second search provider and holding a second key for it.
The whole server is one tool with one parameter, which makes it the cheapest way to get a model off its training data and onto today's web. It returns search results, not page bodies — pairing it with something that fetches a URL is the obvious next step.
What it is
A single-tool server: google_search takes a query string, runs a web search through Google Search via the Gemini API, and returns the results.
What you get
- Web search from one query parameter, answered by Google Search through the Gemini API
- One key for both the model and the search — GEMINI_API_KEY is the only credential it reads
- One entry in the client's tool list, which is the entire integration surface to review
Requirements
A Gemini API key in GEMINI_API_KEY, and npx on your PATH.
Setup effort
One command plus a key — npm install -g mcp-gemini-google-search, then supply credentials