The agent successfully understood Immersive Commons by retrieving machine-readable documentation files (.txt, .md, .json) that contained the core information needed to answer all four questions (what they do, who it's for, pricing, differentiation). The site is highly navigable for agents that know to look for these files, but the main HTML pages are React-rendered shells with minimal content—the agent had to discover and fetch specialized developer documentation rather than assembling answers from the public website.

• ›Step [7] (llms.txt) and step [13] (llms-full.txt) were the primary content sources, providing a concise and comprehensive overview of what Immersive Commons does, its tools, and authentication model. These files explicitly describe the 183-tool surface, the five membership tiers, and the member-run structure.
• ›Step [11] (pricing.md) directly answered the pricing question with a declarative statement: 'There is nothing to buy on this site... every tier upgrade are **free**.' This was the most explicit source for that section of the answer.
• ›Steps [12] (auth.md) and [15] (openapi.json) provided technical depth on authentication mechanisms and API capabilities, enabling the agent to explain how the system is agent-accessible and why it differs from typical SaaS products.
• ›The site surfaces critical information via machine-readable files (.txt, .md, .json at predictable paths like /llms.txt, /pricing.md, /auth.md, /openapi.json) rather than through the main HTML pages. Fetches of /about, /features, /pricing, /join all returned 404 or empty React shells, forcing the agent to rely on prior knowledge that these documentation files exist.
• ›The agent had to guess the existence of pricing.md, auth.md, and llms-full.txt—these were not discovered by following links in the HTML pages but by recognizing a developer documentation pattern. This makes the site agent-ready for agents with prior knowledge of documentation conventions, but not self-discoverable from the homepage alone.
• ›The /members and /projects pages returned HTML but were JavaScript-rendered, making their actual content inaccessible to the agent. The agent acknowledged this limitation in its final response ('the website is mostly JavaScript-rendered').
• ›The site explicitly publishes /llms.txt and /openapi.json, which are standard AI agent discovery files (referenced in Model Context Protocol and OpenAPI standards), indicating intentional agent-readiness at the documentation layer.