Free · ~60 seconds · no signup

See what AI agents understand about your site

Enter your web address below. NL-Agent fetches your site, looks at it the way an AI assistant does and runs 16 checks — from hosting and robots.txt to MCP and DNS. You watch it think out loud and get the full report right after.

We only read your site — nothing is changed, no email needed.

16 checks 0–100 score Level 1–5 on the Cloudflare standard ±60s and you see everything

What happens after you click

Four phases, thinking out loud

No blank loading screen. The scanner tells you at every step what it fetches, what it finds and why it matters.

PHASE 1 · 0–15%

Who are you and where do you live?

We fetch your homepage and look at what runs under the hood.

IP address, country and provider
Web server and CDN
CMS/platform and tech stack
SSL certificate and response time

PHASE 2 · 15–40%

May and can an AI find you?

The first files every AI crawler requests, in the same order.

robots.txt + AI bot rules
sitemap.xml
ARD manifest (ai-catalog.json)
Link headers (RFC 8288)

PHASE 3 · 40–65%

Does the AI understand it?

A model prefers clean text and data over your CSS and JavaScript.

Markdown negotiation
Schema.org / JSON-LD
auth.md (agent_auth)
Titles, meta and heading structure

PHASE 4 · 65–100%

Can an agent actually act?

The difference between a brochure and a usable tool — plus the edge scan.

Agent Card, API catalog
MCP Server Card, Agent Skills
OAuth, WebMCP, Web Bot Auth
DNS-AID + DNSSEC

The 16 signals

Exactly what we check

These are the signals ChatGPT, Claude, Perplexity, Gemini and autonomous agents use to find, read and use your site. Every check gets a verdict, an explanation of why it matters and the concrete fix.

Discoverability

4 checks

robots.txt + AI bot rules: The first place every agent reads. Without explicit rules and Content Signals they guess — and often skip you.
sitemap.xml: The full map of your pages. Without it, deep pages stay invisible.
Link headers (RFC 8288): Point agents straight at your capabilities without parsing your HTML.
ARD manifest (ai-catalog.json): Agentic Resource Discovery: one machine-readable catalog of your MCP servers, agents and APIs. New in the Cloudflare scan (2026).

Content & structure

3 checks

Markdown negotiation: If an agent asks for plain text your server should serve it — now it gets full HTML and understands less.
auth.md: Explains how agents may authenticate. The agent_auth block must sit in a code fence or it is ignored.
Schema.org / JSON-LD: How Google and AI show rich results: prices, ratings, FAQ. Without it you compete one-handed.

Capabilities · MCP & Skills

7 checks

Agent Card (Google A2A): Your identity card for agent-to-agent traffic.
API catalog (RFC 9727): The menu of everything an agent can call.
MCP Server Card: Lets an agent act, not just read: quotes, bookings, orders.
Agent Skills: Which concrete tasks you offer and how an agent calls them.
OAuth (server + resource): How an agent proves who it acts for. Without OAuth agents can only look.
WebMCP: Registers tools in the page itself so an agent can act inside the browser.

Bot access & DNS

2 checks

Web Bot Auth: Lets real agents identify cryptographically so you can block fakes.
DNS-AID + DNSSEC: Agents discover your capabilities at DNS level: the fastest, most reliable route.

Your result

What is on your screen after ~60 seconds

◎

Score 0–100 + your Level

One number for your agent-readiness plus the matching Cloudflare level, from Invisible to Agent-Native.

▤

16 checks, expanded

Per check: pass, partial or fail — with exactly what we found, why it matters and the fix.

◍

Recon of your hosting

IP, provider, country, server, CMS, SSL and response time — the context your site runs in.

⚑

Technical priorities

The fixes that lift your site fastest, sorted by impact instead of alphabetically.

◔

Shareable report link

Your report gets its own URL you can keep or forward to your web developer.

✎

No sales funnel

Everything is visible immediately, ungated. You can hire us to implement it — but you do not have to.

LEVEL 1Invisible — agents barely see you

LEVEL 2Discoverable — you get found

LEVEL 3Readable — you get understood

LEVEL 4Controlled — you set the access

LEVEL 5Agent-Native — agents can act

Before you start

Short answers

Does anything change on my site?

No. The scan is fully read-only: we request the same public pages and files an AI crawler would and write nothing.

What does it cost and do I need to sign up?

Nothing and no. No email, no account, no credit card. You see the full report immediately.

How long does it take?

Usually 40 to 90 seconds. The last step is an edge scan that needs 30–60 seconds itself; meanwhile you see every intermediate result arrive.

Does this work on Shopify, Wix or WordPress?

The scan works on any reachable website. How far you can get afterwards depends on the platform — each fix in your report is tagged (DIY, CMS or edge).

What do you store?

Only the scanned address, the timestamp and the outcome, so your report link keeps working. No personal data.