One endpoint.
Every URL.
One HTTP call against the index, Bearer auth, AIDocument JSON out, the same shape every page on the web resolves to. Pick your starting point: agent developers head to the quickstart, site owners head to the worker.
Quickstart
Get an API key, POST /v1/aidocument with a URL, parse the response. curl, Python, and Node samples that work as-is. 5 minutes from signup to your first AIDocument.
Read the quickstart →Site-owner worker
Optional worker for verified domain owners. Reports AI agent visits from your CDN edge to Lyrenth (Cloudflare and Vercel templates), so Lyrenth sees the full picture of who is reading your site.
Install a worker →Integrations: MCP, SDKs, reader
Plug Lyrenth in as your agent's web reader: an MCP server (npx -y lyrenth-mcp), a one-line GET /v1/read, and SDKs for Python and TypeScript.
Browse integrations →Deeper reading
The contract surfaces below are how an AI agent reads Lyrenth without an SDK.
AIDocument shape
The canonical v2 JSON envelope every response returns. Stable contract, forever.
JSON Schema (draft-07)
Machine-readable contract for the v2 envelope. Validate, generate types.
Short markdown index
The llmstxt.org convention. Five lines telling an agent where to read more.
JSON manifest
Capabilities, endpoints, billing model, bot identification. CORS-permissive.
Crawler identification
User-Agent string, robots.txt rules, ownership verification.
Pricing model
Per-request billing, what counts toward quota, free tier limits.
Get a key, resolve a URL.
Free tier ships with 2,000 AIDocuments / month, no credit card.