The agent assembled a comprehensive explanation of Locus by combining fragments from the site's HTML with external sources. The homepage and pricing page returned only boilerplate React HTML without rendered content; the agent relied heavily on prior knowledge and web search results (13 steps were searches or prior knowledge vs. direct site fetches) to reconstruct what Locus does, who it's for, pricing structure, and competitive positioning. The site itself is not machine-readable—it requires JavaScript rendering to expose its content, making it difficult for agents to extract answers natively.

• ›Steps [1], [5], [6], [7], [9], [10] fetched the site's pages but returned truncated React boilerplate without rendered content; none of these steps directly supplied the final answer. The agent had to abandon these and pivot to web search and prior knowledge.
• ›The agent's actual sources were web search results (steps [14]–[26]), which surfaced third-party summaries, Y Combinator profiles, and competitive intelligence. The site's own pages did not self-document in a fetchable form; the agent reconstructed the narrative from external citations.
• ›The site's primary limitation: it is a client-rendered React app with no pre-rendered HTML fallback, no structured data (schema.org), and no plain-text API documentation published at /docs (only a docs.paywithlocus.com Mintlify instance was reachable). This forces agents to either execute JavaScript or rely entirely on external sources.
• ›Critical information gaps persisted despite the comprehensive final answer: actual per-call pricing rates for specific APIs were not exposed, the relationship between 'Locus Pro' and 'Pay with Locus' remained ambiguous, and detailed competitive positioning was inferred rather than stated on the site.