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. Read it from any language, plug it into the tools you already use, or hand it to your agent as an MCP server.
First call, then the reference
Two pages get you from nothing to a working integration.
Quickstart
Get a key, POST /v1/aidocument with a URL, read the response. Working samples in curl, Python, TypeScript, Go, Java, C#, PHP, Ruby and Rust. Five minutes from signup to your first AIDocument.
Open →API reference
Every endpoint a key can call, with request fields and types, response shapes, the error envelope, the status and error codes worth branching on, and how the rate limit and quota behave.
Open →Your own code
A package where we publish one, plain HTTP everywhere else.
SDKs and languages
Dependency-free clients for Python and TypeScript, a one-line authenticated GET that returns Markdown, and the list of every package we publish.
Open →The AIDocument shape
The grouped JSON envelope every read returns: eight top-level groups, the same shape for every URL and every agent, versioned forever.
Open →No integration code at all
Connect an MCP client, import a framework package, or deploy a template.
MCP server
A hosted endpoint your client connects to with a header, or the published package run locally with npx. Three tools: read one page, read up to twenty, check your usage.
Open →Frameworks and tools
LangChain and LlamaIndex packages, a ready-made tool for the Vercel AI SDK, a one-line tool for eve, and a tool-calling recipe for the OpenAI SDK.
Open →Edge templates
Cloudflare Worker and Vercel Edge Middleware templates that report AI agent visits from your own edge, with the deploy steps for each.
Open →If agents are reading your site
See who reads you, and author what they get.
Site-owner worker
Optional worker for verified domain owners. Reports AI agent visits from your CDN edge, so you see the full picture of who is reading your site and not only the readers who arrive through Lyrenth.
Open →Crawler policy
How our crawler behaves: robots.txt on every fetch, the machine-readable reservations we honour, and how to verify a domain to author what agents read.
Open →Contracts, not prose
How an AI agent reads Lyrenth without an SDK, and without this page.
JSON Schema (draft-07)
Machine-readable contract for the v2 envelope. Validate, generate types.
The agent doing-file
What this service is, how an agent gets access, and exactly which calls exist.
Short markdown index
The llmstxt.org convention. A few 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.