fastCRW

fastCRW lets Claude search the web and read public web pages. Search returns ranked results with titles, URLs and snippets. Reading a page returns its main content as clean markdown instead of raw HTML, with the source URL attached, so answers stay grounded and the context window stays small. Tools: - Web search: ranked results for a query, optionally with each result's content. - Read a page: one URL to clean markdown, HTML or a list of links. - Map a site: list the URLs a site publishes in its sitemap and links, without fetching content. - Crawl a section: read a bounded set of pages under one URL (default 10 pages), as an async job with a status check. - Extract fields: return structured JSON from public pages against a schema you provide. Typical uses are reading documentation to answer an integration question, gathering facts with source links from primary sources, and turning a public page into a table. The tools only read: they never post, submit forms, sign in to third-party sites or change anything outside fastCRW. fastCRW is open source under AGPL-3.0, and the same engine can be self-hosted. A free fastCRW account is required; no card.