Lighthouse · Agentic Browsing

See your site the way
an agent sees it.

Agents don’t browse like people. They read your llms.txt, parse your structure, and call your tools. Paste a URL — agent-ready grades what they actually see, 0–100, then writes the fixes.

Try

Why it matters

An agent reads three things first

/llms.txt

A token-cheap map of your pages and actions — the file agents look for before they spend tokens crawling HTML. Most sites don’t have one.

the structure

Landmarks, an h1, a title, structured data. Without them an agent is guessing at what your page means and where things are.

modelContext

WebMCP tools let an agent act — search, sign up, add to cart — by calling a function instead of reverse-engineering your DOM.

The scoring model

Nine weighted checks, summing to 100

The same checks the CLI and GitHub Action run. The score reflects the server-rendered HTML an agent first sees.

llms.txt22A token-cheap map of your pages an agent reads first.
WebMCP tools22Callable tools so an agent can act, not just read.
Structured data13JSON-LD / OpenGraph meaning machines can parse.
Semantic structure13Landmarks and an H1 so the layout is legible.
Title + description10What the page is, in one machine-readable line.
robots + sitemap8A discoverable, crawlable surface.
Canonical URL4One authoritative address, no duplicates.
Document language4Declares the language of the content.
Image alt-text4Describes the visuals an agent can’t see.

Ship the fix

Two ways to make it agent-ready

Do it yourself — free

agent-ready is the only tool that doesn’t just scan — it opens the fix as a pull request. It detects your framework (Next.js, Vite, static), writes the files, injects the missing tags, and shows a before → after score.

npx agent-ready fix . --pr

Or gate every PR with the GitHub Action:

- uses: VeldinS/agent-ready@v0
  with:
    url: https://yoursite.com
    comment: true

Read the docs

Done for you

Have me make it agent-ready

Want it handled end-to-end? I’ll wire WebMCP tools to your real endpoints, write a proper llms.txt, add structured data, and put a CI gate on it — so it stays agent-ready. Fixed scope, from $1K.

WebMCP tools wired to your actual API
Hand-tuned llms.txt + structured data
CI gate so the score can’t regress

Get it done →