Skip to main content
Crawl follows links from a start URL and returns Markdown for up to 500 pages. Bound the work with page, depth, path, and time limits.

Crawl a site

Read each item in results for its URL, Markdown, and page metadata; the run-level metadata counters and cache_metadata are described on the same page. The reference lists the full response. Check partial when a deadline may have ended the crawl early.

Choose the workflow

Freshness and waits

maxAgeMs is the oldest cached page content the crawl accepts; 0 fetches fresh. waitForMs is the browser wait after initial page load for each crawled page. Requests also accept policies for access policies every URL must pass and tags for usage tracking.

Scope and failures

See scope and limits to restrict discovery and page content to control extraction. A missing start page returns 404 NOT_FOUND; unsupported start-page content returns 415 UNSUPPORTED_CONTENT. Crawl uses one concurrency slot like other calls; on per-minute plans it has a weight of 10 in the API key’s rate-limit bucket. For a job that should continue beyond a single request, use a batch.