Contrie

Contrie turns a public web page into the data you ask for. Describe the fields in plain language or pass a JSON Schema; Contrie fetches the page, extracts the values and returns JSON with an acceptance decision you can act on: a quality score, schema validity, and a per-field check of which values appear on the page, with the supporting excerpt. Every accepted answer carries an Ed25519 signature, so it can be checked later against what Contrie returned.

What you can ask Claude to do with Contrie:

• Extract structured data from a web page: prices, dates, rates, tables, specifications (contrie_extract).
• Read a web page as clean, structure-preserving markdown (contrie_read).
• Search Contrie's curated index of public technical reference pages (HTTP, robots.txt, MCP, bot verification standards) for dated, citable snippets with their source URL (contrie_search).
• Verify data you already have against its source page, field by field, at no credit cost (contrie_verify).
• Watch a page for changes to an answer, read what changed, and list or stop your monitors (contrie_watch, contrie_changes, contrie_monitors, contrie_unwatch).
• Check your remaining credits (contrie_account).

Sign in with a Contrie account the first time Claude needs one; you can create the account during sign-in. The Free plan includes up to 200 credits a month. A read costs 1 credit; an extraction costs 1, 3, 8 or 15 credits depending on the request, plus 2 when the page has to be rendered in a browser; a search that returns results costs 1 credit, and one with no results costs 0. One page per call: Contrie is not a crawler or an open-web search engine (search covers only its curated index), and it reads web pages, not PDFs or other files. Its source check confirms that a value appears on the page, not that the page states it about that field.