llms.txt and Helping Agents Discover Your API

Agents need a way to discover a site's capabilities without guessing. llms.txt is a proposed convention for exactly that: a plain-text file at the site root describing the site and listing the URLs that matter, each with a one-line description.

What goes in it

• A short paragraph about what the site is
• A list of key pages and endpoints, each with a human-readable description
• Optional sections for documentation, datasets, and feeds

Why it helps agents

An agent that reads /llms.txt in one request learns: the API endpoints, the formats offered, the rate limit, the static file paths, and where the blog and RSS live. That is a complete map of the site in under 2 KB of plain text — cheaper and more reliable than crawling.

How YJTOON uses it

YJTOON's /llms.txt points agents at the catalog (categories.json), the endpoints (/api/), the format documentation, the dataset index, and the RSS feed. Every line is a real URL; every description is accurate.

Is it standardized?

It is a proposal, not a W3C standard, and adoption is growing. Treat it as progressive enhancement: it costs one small file, breaks nothing, and gives capable agents a fast on-ramp.