Integrations

Every way to plug Lyrenth in as your agent's web reader. One key works across all of them, and each reads through the same standing index and returns the same clean AIDocument.

from lyrenth import Lyrenth

client = Lyrenth()                    # reads LYRENTH_API_KEY
doc = client.read("https://example.com/post")

print(doc.title)
print(doc.word_count, "words")
print(doc.markdown[:400])             # the cleaned body
# doc.raw is the full AIDocument envelope

# Up to 20 URLs in one call, each failure isolated.
for r in client.read_batch(["https://example.com/a", "https://example.com/b"]):
    print(r.url, r.ok, r.error)

# Or just the text.
markdown = client.read_markdown("https://example.com/post")

import { Lyrenth } from "lyrenth";

const lyrenth = new Lyrenth();        // reads LYRENTH_API_KEY
const doc = await lyrenth.read("https://example.com/post");

console.log(doc.title);
console.log(doc.wordCount, "words");
console.log(doc.markdown.slice(0, 400));   // the cleaned body
// doc.raw is the full AIDocument envelope

// Up to 20 URLs in one call, each failure isolated.
for (const r of await lyrenth.readBatch([
  "https://example.com/a",
  "https://example.com/b",
])) {
  console.log(r.url, r.ok, r.error);
}

// Or just the text.
const markdown = await lyrenth.readMarkdown("https://example.com/post");

The document you get back is flat and ready to use: title, description, markdown, the word count, and the canonical URL. When you want the whole grouped envelope, it is on raw. Errors raise a typed LyrenthError carrying the HTTP status, so you can branch on it.

curl "https://api.lyrenth.com/v1/read?url=https://example.com/post" \
  -H "Authorization: Bearer $LYRENTH_API_KEY"

Add &fresh=1 to force a live re-fetch, or &max_tokens=4000 to cap the body to a context budget. The full parameter list is in the API reference.

Both do the same job. They watch the User-Agent on incoming requests, buffer anything that looks like an AI bot, and post aggregated counts to /v1/ingest/bot/<your-domain> after the response has already gone out, so your visitors never wait on it. Both are wrapped in try/catch at every step: if Lyrenth is unreachable, the failure is that no new data appears in your dashboard, never that your site is down. Neither one reads a request body, a response body, a cookie, or a visitor IP, and the query string is stripped from the path before anything is sent.

Both need your domain verified first, and the per-domain ingest token from /sites/<your-domain>/settings. Tokens start with aiwt_ and are not API keys: one can only post bot counts for its own domain.

The worker page has the code for both, along with what is collected and what is not. Read the site-owner worker page.

The MCP server is also listed on Claude directory, Cursor directory, Smithery, Glama, LobeHub and GitHub.