Skip to main content
Scrape reads one URL and returns it in the formats you ask for.
Without formats you get the page’s main content as Markdown. Send 1 to 20 formats. A read gives up after timeout: 60,000 ms by default, at most 300,000. Calls without a key have lower limits and fewer fields, see Without a key.
data has one field per format. data.metadata describes the page, see Page metadata. To read the result again, send GET /v1/scrape/{id} with metadata.scrapeId.

Formats

A format is a name, or an object with its fields: { "type": "highlights", "query": "refund policy" }. summary, highlights, question, product, menu and json with schema or prompt use a model and add a fee, see Scrape add-ons.

Options

Every option sits next to url. The API reference has every field.

Errors and warnings

  • A page that answers 404 or 410 comes back with that metadata.statusCode and a metadata.error, at no cost.
  • A problem that does not fail the call adds a warning: a browser action failed, maxChars cut the text, or onlyCleanContent or checkPromptInjection did not finish.
Every other code is in Errors.

Next

  • Interact: keep working on the page in a browser.
  • Map: list a site’s URLs.
  • Crawl: read a whole site.
  • Batches: read a list of URLs.