Astro Removed Its Llms.txt (dacharycarey.com)

Astro, a prominent web framework, has made a significant shift by removing its llms.txt files, which previously served as a structured resource for AI agents navigating its documentation. This change, recorded in a recent pull request, lowered Astro's documentation-compatibility score from a C to an F. The decision was justified by low traffic data on llms.txt files and the desire to optimize build times, shifting focus instead to the new MCP server for agent interactions. The reasoning raises concerns, however, as it overlooks the potential for llms.txt as a reliable navigation layer for various AI agents, particularly those that do not support or require the configuration of the MCP server. The implications of this decision are substantial for both documentation visibility and agent usability. Without llms.txt, agents now face challenges in discovering relevant content, relying on outdated training data instead of an authoritative index. The MCP server, while a more sophisticated solution, caters to a narrower audience, including only those familiar with its setup. This disparity could hinder less experienced developers from efficiently utilizing Astro's documentation. Thus, while Astro's shift may reflect an engineering decision aimed at optimizing resources, it risks neglecting broader accessibility for the growing array of AI tools and workflows in the rapidly evolving documentation landscape.