Watch competitor sites with Firecrawl
Connect Firecrawl securely and monitor meaningful public-site changes on a schedule.
The Competitor site watch template uses Firecrawl search and page scraping to report meaningful public changes in pricing, packaging, positioning, proof, and product claims.
Connect Firecrawl
An operator must first enable MANAGED_MCP_FIRECRAWL_ENABLED. Open
Connections → Firecrawl and prefer OAuth. Browser authorization uses
Firecrawl's dedicated https://mcp.firecrawl.dev/v2/mcp-oauth endpoint.
For unattended setup, use a dedicated Firecrawl API key. Work on Repeat sends
it as Authorization: Bearer … to https://mcp.firecrawl.dev/v2/mcp, encrypts
it at rest, and never places it in the endpoint URL.
Do not put the key in the URL
Firecrawl's current contract explicitly requires a bearer header. Legacy key-in-path URLs are obsolete. Work on Repeat redacts and encrypts legacy secret-bearing MCP targets for migration safety, but new Firecrawl connections never create one.
Reviewed tool boundary
This release permits exact schema matches for two read tools captured from hosted Firecrawl server 3.24.0:
firecrawl_search— find current public sources with bounded domain, category, result, and optional scrape settings.firecrawl_scrape— retrieve one supplied public HTTPS page in a selected format.
Crawl, map, agent, monitor, interact, research, parse, and every other account tool remain blocked. Firecrawl schema drift fails closed. Work on Repeat also blocks private/local URL literals, URL credentials, TLS verification bypasses, and saved browser profiles.
Configure the routine
Choose Competitor site watch, enter the public pages to monitor, attach Firecrawl, and prove one manual run before scheduling it. The routine should compare against prior findings, quote the exact wording behind a material change, cite the URL and retrieval date, and say explicitly when nothing meaningful changed.
Firecrawl reads open-world third-party content, which can contain prompt injection or stale claims. Treat page text as evidence, never as instructions; prefer official primary sources and separate facts from inference. Public availability does not make sensitive personal data appropriate to collect.
Search and scrape consume the connected Firecrawl account's current credits and rate limits. Keep result counts, formats, and page scope narrow. If OAuth is revoked or a key is rotated, reauthorize or replace the key on the saved connection in place.
See Firecrawl MCP documentation for current authentication, free-tier, plan, and tool availability.