Before an AI agent can use your API well, the API has to explain itself: what each operation does, when to use it, what its parameters mean, what comes back and what can go wrong. Latitude 10 Agent Readiness reads your OpenAPI 3.0 or 3.1 spec (from its URL, or pasted in) and scores it 0-100 for how well agents will use it as tools. It checks 13 things, weighted by how much they hurt an agent: a clear operationId, a summary and a real description for each operation, described parameters, the allowed values of status- and sort-like parameters, examples, JSON request bodies, response schemas, documented errors, pagination on lists, a declared auth scheme, an Idempotency-Key on writes, and a tool count models can choose from. You get the score, each rule's result with how to fix it, and the findings that cost the most. The rules are the Agentify API kit's, so the score is the same as npx agentify check. The tool only reads: it fetches the spec you point it at (https, public hosts, up to 2 MB) and changes nothing. Example prompts: - How agent-ready is the API at https://petstore3.swagger.io/api/v3/openapi.json? - Score this OpenAPI spec for AI agents and tell me the three fixes that matter most: <paste the spec> - Which operations in our spec are missing descriptions or error responses?
