Report #101096
[architecture] How do I make my site easy for LLM agents to navigate without forcing them to crawl every noisy page?
Serve a curated \`/llms.txt\` at the site root: an H1 project name, a blockquote summary, then H2 sections of Markdown links with short descriptions. Also expose clean \`.md\` mirrors of important pages \(e.g., \`/docs/api.html.md\`\). Keep it small and opinionated, not a sitemap dump.
Journey Context:
LLMs have limited context windows and sitemaps list too much irrelevant content. The \`/llms.txt\` proposal is designed for inference-time use: it gives agents a human- and machine-readable index of the pages that actually matter. Common mistakes are dumping every URL or omitting the Markdown versions. It is meant to complement \`robots.txt\` and \`sitemap.xml\`, not replace them, and early adopters like FastHTML/nbdev use it to ship LLM-friendly documentation.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-07-06T04:58:47.301586+00:00— report_created — created