# Lyrenth, for agents

Lyrenth is the AI-readable web index: an independent index of over 2
billion public web pages, each served as one clean AIDocument (markdown,
structure, provenance, token economics) instead of raw HTML. Reads are
typically 90 to 97 percent smaller than the original page.

## Get access

1. A human signs up at https://www.lyrenth.com/signup (free tier, 2,000 AIDocuments a
   month, no card) and creates an API key at https://www.lyrenth.com/dashboard/keys.
2. Send the key as: Authorization: Bearer <key>

## The API surface for keys

Keys call the documented product endpoints and nothing else; account
management is dashboard-only by design, so a leaked key can spend
credits and nothing more.

- POST https://api.lyrenth.com/v1/aidocument        body {"url":"..."}; one clean AIDocument
- POST https://api.lyrenth.com/v1/aidocument/batch  up to 20 URLs, per-URL error isolation
- GET  https://api.lyrenth.com/v1/read?url=...      the same read as plain markdown
- POST https://api.lyrenth.com/v1/submit            queue a URL for indexing, free
- GET  https://api.lyrenth.com/v1/quota             your plan and remaining credits
- GET  https://api.lyrenth.com/v1/stats             public index counters, no key needed

Failed fetches never count against your quota. Cached reads answer in
well under a second; a first-time URL is fetched once, then everyone
gets the fast path. A query endpoint (question in, best-matching
AIDocuments out) is built and arrives with the search launch.

## MCP and SDKs

- Hosted MCP server: https://api.lyrenth.com/mcp, streamable HTTP, stateless. Send the
  same Authorization: Bearer <key> on each request. Nothing to install.
- Local MCP server: npx -y lyrenth-mcp (npm; also in Claude's
  connector directory, the Cursor directory, and on the official MCP
  registry as com.lyrenth/lyrenth-mcp, Smithery, Glama and LobeHub)
- SDKs: npm "lyrenth" and PyPI "lyrenth", plus PyPI
  "langchain-lyrenth" and "llama-index-readers-lyrenth"
- Full docs: https://www.lyrenth.com/docs/api, https://www.lyrenth.com/docs/mcp, https://www.lyrenth.com/docs/frameworks

## Conduct

Lyrenth respects robots.txt on every fetch, identifies its requests
cryptographically, and does not train foundation models on crawled
content. Site owners: https://www.lyrenth.com/crawler-policy explains our crawler, and
verified owners get bot analytics and controls at https://www.lyrenth.com/sites.

More for reading agents: https://www.lyrenth.com/llms.txt and https://www.lyrenth.com/llms-full.txt
Contact: hello@lyrenth.com