Run agent-driven structured extraction across many pages with Firecrawl Extract, then poll for results (POST /v2/extract)

Documented steps

1. AUTH: Authorization: Bearer fc-YOUR_API_KEY on both the start POST and poll GET
2. POST https://api.firecrawl.dev/v2/extract with body {"urls":["https://example.com/products/*"],"prompt":"Extract the product name, price, and SKU from each page"}
3. For a strict output shape add a JSON Schema: {"urls":[...],"schema":{"type":"object","properties":{"name":{"type":"string"},"price":{"type":"number"}},"required":["name","price"]}}
4. urls accepts glob patterns (trailing /*) so one call can fan out across a whole site section
5. Optional flags: enableWebSearch (agent may pull supplementary pages), includeSubdomains (default true), showSources (per-field source URLs)
6. Start response: {"success":true,"id":"<extract-uuid>","invalidURLs":null|[...]} — store the id
7. Poll: GET https://api.firecrawl.dev/v2/extract/{id}; status is processing | completed | failed | cancelled
8. On completed, read data (the extracted objects) and tokensUsed (LLM cost accounting)
9. Alternative to polling: pass a webhook config in the start body to get extract.started/completed/failed events