The agent successfully assembled a comprehensive overview of HeyGen's developer platform by combining direct fetches from the documentation site with the company's machine-readable index files. The site proved highly navigable for structured discovery—llms.txt and llms-full.txt directly indexed all major features, pricing, and capabilities—but pricing details and feature comparisons required parsing markdown docs and inferring from fragments rather than being exposed on a single, comprehensive page. The agent filled gaps about competitive positioning using external web search, not the site itself.

• ›Steps [6], [7], [8], [9] returned structured markdown documentation (pricing, quick start, API key setup) that directly answered the 'what they do' and 'how it's priced' questions. The llms.txt file ([6]) was especially agent-friendly, providing a concise machine-readable index with links to all major product surfaces.
• ›Steps [1], [14], [15] established the broader business context (85,000+ users, 175+ languages, enterprise positioning) via homepage metadata and the main product site's llms.txt. The developer portal homepage and marketing site index files were the discovery mechanism—the agent did not need to search or guess to find these.
• ›Pricing and differentiation were NOT cleanly surfaced on the developer portal itself: the /pricing endpoint returned 404 ([4]), forcing the agent to extract per-operation costs from markdown fragments ([7]). Enterprise pricing was noted as custom/unpublished. Competitive differentiation came entirely from external web search ([13]), not HeyGen's own docs—the site does not self-position against alternatives.
• ›The site's agent-readiness is high for structured data (llms.txt, markdown API docs, clear authentication/quick-start flows) but incomplete for business intelligence (no pricing comparison table, no competitor positioning, no feature capability matrix visible to fetches).