The agent successfully assembled a comprehensive explanation of Firecrawl by combining the official site (www.firecrawl.dev homepage, pricing page, and docs) with third-party review and pricing articles from web search. The official site was navigable but required JavaScript rendering; the agent relied heavily on external sources and prior knowledge to fill gaps around pricing nuance, competitive positioning, and feature limitations that weren't clearly documented on the primary domain.

• ›Step [4] (firecrawl.dev homepage) was content-bearing but heavily JavaScript-dependent; the agent extracted high-level positioning ('AI Web Crawler Built for LLM Applications') but had to supplement with external sources for detailed feature breakdowns and pricing tables.
• ›Steps [8] and [9] (web searches for pricing and alternatives) were explicitly cited and provided concrete pricing tiers, cost comparisons, and competitive positioning that the official site did not surface clearly in a single, machine-readable format.
• ›Step [5] (pricing page) was fetched but its raw HTML was not extractable in the final response—the agent instead cited external pricing aggregator sites (eesel.ai, costbench.com) that had parsed and reformatted the pricing information, indicating the official pricing page was not agent-friendly.
• ›Step [6] (docs.firecrawl.dev) was fetched but the agent noted it required digging and was not the primary source for the final answer; the docs exist but are not SEO'd or indexed as a discovery surface.
• ›The agent explicitly called out unresolved gaps (data residency, SLAs, feature breakdown, open-source scope, rate limits for paid tiers) that suggest the official site does not publish this information in accessible form, even after traversing homepage, pricing, and docs.
• ›The domain firecrawl.ai (step [1]) was parked/inactive, forcing the agent to search for the correct domain (firecrawl.dev); this adds friction to cold discovery.