Extract

Extract

Public HTML as clean text, when publisher and network policy allow it.

Live contract — August 11, 2026: Extract accepts public HTML URLs only, up to 2,048 characters. It checks the index first. A live fetch on a miss is gated: anonymous callers and free browser JWTs stay index-only; Growth plan JWTs and API keys may live-fetch. Robots rules, noindex, nosnippet, and private-network protections can refuse a URL.

What it does

One tool: extract_url(url). It returns title, text, and passages (paragraph blocks), plus a source field that tells you where the content came from:

  • "index" — we already crawled this page. The text is byte-identical to what search_web quotes and carries its crawl date, so the page your agent cited is the page it can pull verbatim. No network call is made.
  • "live" — we hadn't crawled it, so it was fetched on the spot. Nothing is stored. Live fetch requires Growth (browser) or an API key — not the free/anon tier.

What it honors

We respect the publisher's own directives on both paths: a noindex page is refused outright, and a nosnippet page returns its title and link with empty text. That's a feature — a page we're not allowed to serve comes back as an explicit refusal, never as quietly-degraded content.

Why the limit is tighter

Extract runs at 10 calls per minute — tighter than search on purpose. An index hit is one local lookup, but a miss makes SeaWeb fetch someone else's page on your say-so.