Player not loading? Watch on YouTube
Firecrawl's Scrape and Parse endpoints accept documents in different ways: Scrape takes a reachable URL, while Parse takes a file on disk. The presenter demonstrates PDF extraction through Claude Code with a Firecrawl parse skill and mentions language-specific SDKs. Downloading an online file with curl is another route to Parse; Scrape handles the URL directly, with retries and redirects.
The comparison focuses on batch processing, document size and browser actions. According to the presenter, Parse requires separate calls per file and your own retry handling. Batch Scrape provides managed jobs with polling and timeouts. The speaker states a 50 MB limit for Scrape and batch documents, then describes a larger-file storage path that might not yet be released. Treat that size extension as a prospective feature rather than an available capability.
Scrape also supports clicking, scrolling and waiting on pages, including JavaScript-heavy sites. Both endpoints have PDF options for per-page markdown, page markers, layout blocks and a maximum page count. The presenter says these options add no extra credits, and describes extraction as costing one credit per page in this example: a single URL containing a 300-page PDF can therefore use 300 credits.
The closing section introduces PDF Inspector as a free, open source Rust tool for offline processing. The speaker claims it can parse 200 PDFs locally in under three seconds, without providing benchmark conditions.
