POST /web/scrape with formats.parse: true and parseParams: {"type": "prebuilt", "parser": "links"}. Read the result in parsed.data, with parsed.type: "links". See the Quickstart for authentication and SDK setup.
Don’t send custom rules with this parser. For CSS selectors and rules, use the Custom parser.
Request
Use any supported page URL. The parser reads anchors with anhref on that page; it does not crawl destinations. For site-wide URL discovery, use Map URLs.
Output
Whenparsed.success is true, the parsed output has this shape:
Fields
<base href>, or the final page URL. Only HTTP and HTTPS URLs are returned; fragment-only references, mailto:, tel:, and JavaScript links are excluded.
Links are deduplicated by their resolved URL. The first non-empty label wins: visible text is preferred within an anchor, then aria-label, then image alt text. Missing labels are null. Hidden anchors can still contribute a URL with text: null; anchors inside templates are excluded.
Empty results and errors
No matching links producelinks: []. Read parsed.success before using the data. A failed parse has data: null and can coexist with successful outputs. parsed.type remains "links" even when parsing fails. See timeouts and errors for handling failures.
Request options
This parser runs on page HTML after content filtering. Browser actions can change the page before extraction, and browser profiles can provide an authenticated session. Cached HTML can be reused; setmaxAgeMs: 0 for fresh data. Timeouts and zero data retention also apply.