Half of what agents need to read arrives as a PDF or an office file, not a web page, and the usual answer is a separate parsing dependency. Having scrape, search and document parse behind one connection — with targeted answers and JSON matching as output formats — means the extraction step produces something structured rather than a wall of text. The keyless tier makes it trivial to evaluate.
Firecrawl's hosted endpoint, which listed three tools to our probe as firecrawl-fastmcp 3.24.0. It works keyless within daily limits, which is unusual for a scraping service.
- firecrawl_scrape — extract content from a URL, with format, tag include and exclude, location and freshness controls
- firecrawl_search — web, news or image search with operators, domain filters and highlights
- firecrawl_parse — turn a supported document into markdown, HTML, links, a summary, targeted answers or matched JSON
Nothing to start — keyless with daily limits. A bearer API key or OAuth through their dedicated endpoint lifts them.
Paste a URL, then authorize — add the endpoint to your client, then approve the OAuth consent screen