Public betaUniversal adapterAI-readable web indexindexed pages

The web,
translated for AI agents.

Lyrenth is the universal adapter for the existing web: send any public URL, get clean AIDocument JSON with markdown, headings, links, structured data, crawl trace, and token economics. No scraping stack. No HTML cleanup. No website changes required.

Live AIDocument preview
Try:
LYRENTH IS IN PUBLIC BETA · LOGOS APPEAR HERE AS CUSTOMERS GO PUBLICBe one of them →
Your logo here
Your logo here
Your logo here
Your logo here
Your logo here
Your logo here
Your logo here
Your logo here

Be among the first agents to use Lyrenth. Free tier ships with 2,000 AIDocuments per month, no credit card.

Get an API key
01 · Economics

Source noise taxes every agent.
We return the signal.

A raw page carries scripts, styles, navigation, tracking, and app shell bytes. Lyrenth fetches once, renders when needed, strips the noise, and returns AIDocument JSON that is often 60-95% smaller for model input.

SAMPLE PAGE
Source noisefetch + send to model
Tokens fed to LLM
240,420
Cost (input only · $3 / 1M tok)
$0.7213
What's in the payload
  • <script> bundles
  • <style> blocks
  • Nav, header, footer
  • Cookie banners
  • Tracking pixels
  • The actual content
AIDocument/v1/aidocument
Tokens fed to LLM
2,992
Cost (input only · $3 / 1M tok)
$0.0090
What's in the payload
  • Title + description
  • Headings tree (H1-H6)
  • Markdown body
  • Outbound link graph
  • JSON-LD structured data
  • OpenGraph + meta
Tokens (1px ~ 200 tokens)240,420 raw · 2,992 ours
0
Reduction 98.8%That's 80× cheaper per fetch
240,420
02
nav.menu home
cookie.banner accept
article.title…
script src=
aria-hidden true
"The article
discusses cookies
and the home
menu before…"

Cleaner context, fewer bad answers

When an LLM sees nav menus, cookie banners, and tracking scripts mixed with content, it can confidently summarize the wrong thing. We strip noise so the model receives the page, not the furniture around it.

03
STATIC FETCH
~50ms
HEADLESS FALLBACK
2 to 5s
Static for most pages, headless escalation only when needed.

JavaScript handled.

A meaningful share of modern sites need rendering. We try a fast static fetch first, and only escalate to headless Chromium when the page is empty. Fast first, headless when needed.

04
"economics": { raw_tokens: 240,420 our_tokens: 2,992 savings_usd: 0.7123 }

Cost transparency, every call

Every API response carries an economics block: raw HTML token count, AIDocument token count, and estimated savings at frontier-model input rates. No black-box math. The comparison travels with the document.

02 · One endpoint

Two minutes to your first
clean AIDocument.

HTTP GET, Bearer auth, JSON out. Same shape for every page on the web. No SDK gymnastics, no per-site selectors, no rendering decisions to make.

request.curl
response.json
○ pending
Endpoints
5
aidocument · document · submit · stats · sites
Output shape
1
canonical AIDocument JSON, every URL
Render fallback
auto
static then headless when needed
Public shape
open
AIDocument contract published in /llms-full.txt
03 · For site owners

Lyrenth handles
AI agent traffic for you.

Whether your site is AI-ready or not, we make it AI-readable. Verify your domain and Lyrenth does the work: clean, structured pages served to every AI agent reading you, plus an AI Readiness score built to account for standards like llms.txt, context.txt, structured data, and agent action manifests as support expands.

Free for site owners. Forever.

SAMPLE PREVIEW · ILLUSTRATIVE DATAYour numbers will look like this once you verify a domain.
www.lyrenth.com/sites/yourdomain.com
Verifiedvia DNS TXT · 4 days ago (sample)

yourdomain.com

AI Readiness score
87
GOOD
↑ 12 pts since last month
Static-renderable
94
Cache-friendly
84
Structured data
71
Semantic markup
89
AI requests
38,142
↑ 28% vs prev period
Unique agents
14
GPTBot, ClaudeBot, Perplexity…
Avg response
284 ms
cache hit 84%
Daily AI agent requests
Sample · last 30 days
Cache hit Fresh fetch
Top requesting agents
01GPTBot14,493
02ClaudeBot9,154
03PerplexityBot6,484
04Anthropic-Agent4,196
05GoogleOther2,288
06CCBot1,525
What Lyrenth does for you
Behind the score: the work that happens automatically once you verify your domain.
Resolve pages into one clean AIDocument
We turn your pages into AIDocument JSON: markdown, headings, links, JSON-LD, and AI-readable hints as we add support for them. Every agent fetching through Lyrenth gets the same clean shape.
Render JavaScript when needed
If your page is JS-rendered, Lyrenth automatically escalates to headless Chromium. Agents see content, not an empty SPA shell.
Cache so agents don't hammer you
Identical requests from different agents collapse to a single fetch from your origin. Your servers stay quiet; agents still get current data.
03.1
Verify in 60 seconds

DNS TXT or .well-known file. We query 1.1.1.1, 8.8.8.8, and 9.9.9.9 in parallel, no flaky local resolvers.

_aiwebindex-verify.yourdomain.com TXT aiwi-verify=g7f3...
03.2
Recognizable User-Agent

We identify ourselves with a clear UA so you can see us in your existing logs and tooling. Domain-wide opt-out by email today; verified-owner per-path controls and robots.txt Disallow honoring are on the short-term roadmap.

User-Agent: AIWebIndex/1.0 (+https://lyrenth.com/bot)
04 · Bridge, not protocol
AIDOCUMENT v1.0 · OPEN SHAPE
The contract
{
  url, canonical_url,
  title, description,
  markdown,
  headings[],
  links[], images[],
  meta {
    author, language,
    word_count, reading_time
  },
  structured_data {},
  crawl { render_mode, fetch_ms },
  economics { token_savings_percent }
}

The web will not
change first.
So we translate it.

Billions of pages already exist. Asking every owner to publish a new protocol before AI can read the web is a decade-long adoption bet. Lyrenth takes the opposite path: read the messy public web as it is, then serve one clean shape agents can trust.

We publish AIDocument as the output contract, not a requirement for site owners. Agents get markdown, headings, links, structured data, crawl trace, and economics in one JSON document. Website owners do not need to install anything.

Emerging AI-readable standards still matter. llms.txt, context.txt, agent manifests, structured data, and browser-exposed tool hints can enrich the document. They are bonus signals, not prerequisites.

The same clean index becomes the base for semantic search. That search layer is planned for private beta; the AIDocument API is live today.

01
No adoption required
Lyrenth reads existing public pages as they are. A site can be messy, static, rendered, structured, or unstructured; the output to agents stays clean.
02
One open shape
The AIDocument JSON envelope is published as markdown at /llms-full.txt and as JSON at /api/agent-manifest. Agents can build against the same contract.
03
Standards-aware
We treat llms.txt, context.txt, AI action manifests, WebMCP-style hints, and JSON-LD as inputs. They enrich the document; they are not required for Lyrenth to read the page.
04
Origin-friendly cache
When many agents resolve the same URL within our cache window, the origin site receives one fetch. We absorb read amplification so the open web does not have to.
05 · Pricing

Simple per-request pricing.
Live free tier today.

Every successful API call counts as one request. No surprise bills, no usage tiers nested inside usage tiers. Switch plans anytime, cancel anytime, no annual lock-in.

Free tier is live today. Pro and Custom ship in our next phase.

Free
For getting started
$0forever
Get an API key
  • 2,000 AIDocuments / month
  • AIDocument JSON output
  • Domain verification
  • Community support
Pro
For shipping agents at scale
Coming soonHigher limits + per-key analytics
Get the free tier

Free-tier users get notified when Pro launches.

  • Higher monthly limits
  • AIDocument JSON output
  • Per-key analytics + cost report
  • Domain verification (unlimited)
  • Email support
Custom
Enterprise / on-prem / volume

Coming soon. Volume, on-premise, or SLA conversations once we ship.

[email protected]
Site owner free tierVerify your domain. Lyrenth handles the rest. Free for site owners. We serve AI agents on your behalf.
Verify a domain
06 · Questions

Things people ask
before signing up.

Real answers to real questions. If something here is wrong or unclear, mail us and it will land on a human.

Lyrenth is the AI-readable web index for agents. You give us a URL, we return the same page in a clean structured AIDocument shape: markdown, headings, links, JSON-LD, and a per-call economics report. One endpoint, one response shape, every URL.

Existing web,
AI-readable now.

Build agents on clean web context. Or verify a domain and let Lyrenth serve AI-readable pages on your behalf.