Deep Scrape

Turn one subject and a few known URLs into structured evidence across public sources. DeepAPI discovers relevant sources and returns profiles, posts, people, websites, and conflicting claims in one JSON dossier.

Prepare the request

Read the sibling deepapi skill for credentials, headers, and the shared API protocol. Use POST /v1/scrape/deep. The API key needs scrape:deep scope.

• query is required and limited to 500 characters. Name one subject, add identifying context, and state the information that matters. Keep the longer assignment for your own synthesis.
• urls optionally supplies known public website or profile URLs. Use them to reduce namesake mistakes. They anchor discovery; they do not restrict results to those URLs.
• sources optionally restricts source types, for example ["website", "github", "twitter"]. Omit it for broad discovery. This is not a domain filter, and it can exclude source types needed to use a seed URL.
• maxCostUsd defaults to "0.50", with a "5.00" hard maximum per request. This is a spending ceiling, not a quoted charge. Respect the user's total budget across all calls.
• dryRun: true previews the credit hold without scraping or charging. A preview contains no dossier; the paid call must omit dryRun or set it to false.

Only send documented fields. There are no model, provider, depth, maxItems, outputSchema, or separate instructions controls on this endpoint. Requesting a fact in query does not guarantee it will be found or add a new response field.

If the schema, price, scope, or availability is unclear, fetch GET /v1/capabilities?capability=scrape.deep. Its live contract takes precedence over this file.

Start and keep the request identity