Summary
Every Google surface, one tool each.
A single 'search' tool flattens Google into web results and loses the rest. Splitting it — Scholar for papers, Patents for prior art, Maps and Reviews for a place, Lens for an image — means the agent asks the right index instead of hoping the general one covers it.
What it is
A Python MCP server fronting Serper's Google Search API. Thirteen tools: one per Google surface, plus a page scraper.
What you get
- Standard Google web search
- Image, video and news search
- Shopping results, and Google Lens lookups
- Maps and Places results, and the reviews attached to a place
- Google Scholar and Google Patents search
- Autocomplete suggestions, for seeing what people actually type
- A webpage fetched and scraped, so a result can be read rather than only listed
Requirements
A Serper API key as `SERPER_API_KEY`, and a current Python. Run it with `uvx serper-mcp-server` — install `uv` first — or let Smithery's CLI install it into Claude Desktop for you.
Setup effort
One command plus a key — pip install serper-mcp-server, then supply credentials