Migrating from /extract

Migrate this codebase off the Firecrawl /extract API. POST /v1/extract, POST /v2/extract, GET /v1/extract/{id} and GET /v2/extract/{id} shut down on Thursday, December 10, 2026, 15:00 UTC, and return HTTP 410 after that.

1. Find every /extract call:
   - Raw HTTP requests to /v1/extract or /v2/extract on https://api.firecrawl.dev (or a self-hosted base URL), including the GET /extract/{id} status polls.
   - SDK calls: extract, start_extract, async_extract, get_extract_status (Python, firecrawl-py); extract, startExtract, asyncExtract, getExtractStatus (Node, @mendable/firecrawl-js or firecrawl).
   - MCP configs that pin firecrawl-mcp below 3.23.7, whose firecrawl_extract tool still calls /extract.
2. Upgrade the Firecrawl SDK to its latest release:
   - Python: pip install --upgrade firecrawl-py. Use the Firecrawl class (or AsyncFirecrawl): scrape, batch_scrape, start_batch_scrape, get_batch_scrape_status, agent, start_agent, get_agent_status.
   - Node: npm install @mendable/firecrawl-js@latest (or firecrawl@latest, if that is the package installed). Use the Firecrawl class: scrape, batchScrape, startBatchScrape, getBatchScrapeStatus, agent, startAgent, getAgentStatus.
   - MCP: upgrade firecrawl-mcp to its latest release, or use the hosted server. Use firecrawl_scrape with a JSON format or firecrawl_agent instead of firecrawl_extract.
3. Rewrite each call with this mapping, keeping its behavior the same:
   - One URL without a /* wildcard -> POST /v2/scrape with url and formats: [{"type": "json", "prompt": <prompt>, "schema": <schema>}]. It is synchronous: drop the polling and read the result from data.json (doc.json in the SDKs).
   - Several URLs, one result per page -> POST /v2/batch/scrape with urls and the same json format (inside options in the Node batchScrape call). Poll GET /v2/batch/scrape/{id}. Each item in data has its own json and metadata.sourceURL; merge them in code where the old code expected one object.
   - Several URLs, one combined answer -> POST /v2/agent with urls, prompt, schema and strictConstrainToURLs: true. Poll GET /v2/agent/{id} until status is "completed"; the result is in data.
   - A /* wildcard, enableWebSearch: true, allowExternalLinks: true, or a prompt with no urls -> POST /v2/agent with prompt and schema. For a wildcard, pass the site URL without /* and leave strictConstrainToURLs unset.
   - /extract scraped whole pages unless the call set scrapeOptions, and /v2/scrape defaults to onlyMainContent: true. Where the old call had no scrapeOptions, send onlyMainContent: false on scrape and batch scrape.
   - scrapeOptions.* -> top-level scrape options on /v2/scrape and /v2/batch/scrape. /v2/agent takes none.
   - systemPrompt -> put its text at the start of prompt.
   - timeout -> timeout per page in ms on scrape and batch scrape. /v2/agent has no request timeout: cap it with maxCredits.
   - webhook, threatProtection, origin and integration keep their names. ignoreInvalidURLs stays on batch scrape only.
   - showSources and __experimental_showSources -> metadata.sourceURL on each scraped page.
   - Remove agent (the FIRE-1 option), urlTrace, includeSubdomains, limit, ignoreSitemap and every other __experimental_* option.
   - Status values: batch scrape returns scraping, completed, failed or cancelled; agent returns processing, completed or failed. expiresAt and creditsUsed keep their names.
4. Run the project's tests and fix any failures.

Reference: https://docs.firecrawl.dev/migrating-from-extract.md