Fetch a web page as clean extracted content

URL to fetch

Output format. 'markdown' (default) clean article text; 'html' raw cleaned HTML; 'json' the structured SSR blob (TikTok, Pinterest, etc.) instead of article text.

CSS selector to scope the result. An anchor-id selector (e.g. a docs anchor like '#usage') that lands on a heading expands to the heading's whole section, up to the next same-or-higher-level heading; other selectors return the matched elements themselves.

Article title, when extracted from the page metadata.

Article author/byline, when extracted from the page metadata.

Article publication date (ISO 8601), when extracted from the page metadata.

Last-modified date (ISO 8601), when the page publishes one.

Page summary, from og:description / meta description.

Publishing site name, from og:site_name.

Site section/category, from article:section.

Page language, from the html lang attribute.

Lead image URL, from og:image / twitter:image.

Site favicon URL.

Classified result type. Use this to branch your retry / strategy logic.

True when the URL is paywalled. Set alongside the real outcome (e.g. teaser, not_found, bot_challenge).

When outcome=unsupported_target, why the URL pattern is permanently unsupported.

Full extracted content length in characters, before any inline truncation.

True when content was truncated to fit the inline cap; full content at download_url.

Characters NOT included inline (available via download_url).

Presigned HTTPS URL to download the full extracted content. Set when truncated.

Presigned HTTPS URL to download the ORIGINAL file (PDF, docx, pptx, ...) this content was extracted from. Distinct from download_url, which is the extracted text. Set whenever the fetch parsed a real file - including one we identified but declined to convert, so the bytes are still reachable.

Size of file_url's object in bytes.

Content type of file_url's object, e.g. application/pdf.

ISO 8601 UTC timestamp when download_url / file_url expire (~1h).