Discovery layer
The discovery layer is the machine-legible surface of an API: the entry-point files and structured signals — llms.txt, AGENTS.md, an OpenAPI spec, MCP manifests and registry presence, crawlable sitemaps — that let an agent find a usable version of the documentation on a plain HTTP fetch. It is the gate: an API that fails here loses before its documentation quality is ever tested.
Agents arrive over plain HTTP, without executing JavaScript, and they navigate by machine-readable entry points rather than by browsing. The discovery layer exists so that an arriving agent lands on the API reference instead of a marketing page or an empty JavaScript shell. Each surface plays a role: llms.txt declares the agent-facing entry point, AGENTS.md briefs a coding agent, the spec gives a parseable contract, MCP surfaces announce a tool interface.
Failures at this layer are common and concrete. Some documentation renders essentially nothing without JavaScript; some llms.txt files route agents to marketing pages instead of the reference. These are verifiable properties of the public surface, and they tend to be the cheapest problems on an API's profile to fix — publishing a correct entry-point file is a far smaller job than rewriting documentation.
Discovery is graded first, as its own axis of the Discry Score. Discry checks whether the machine-legible path exists — fetching each signal over plain HTTP, the way agents actually fetch — and serving these surfaces well is rewarded directly. When an API publishes an llms.txt, that file is treated as its declared agent-facing entry point, and if it routes agents to marketing pages the grade reflects the surface the API itself chose to present. Discovery failures publish with the reason and the URLs tried.
What does an AI agent make of your API?
Find out in about a minute — no signup.
Discry your API — free