extract-structured-data takes a prompt and a URL, so the shape of the answer comes from the description written for it rather than from CSS paths that break at the next redesign. That single call is the whole surface: there is no crawl, no pagination and no session handling here, so pages behind a sign-in or spread across many URLs need something else alongside it.
The official Scrapezy server, whose single tool extracts structured data from a web page.
- Turn one URL into structured data shaped by a prompt written in plain language.
- Describe the fields you want rather than the markup they sit in — the call takes a prompt and a URL, and nothing else.
Each call carries the URL it should fetch, and there are no environment variables to set beforehand.
One command plus a key — npx -y @scrapezy/mcp, then supply credentials
