Is your site ready for AI agents?

AI agents do not browse like people. They look for machine-readable contracts: a .well-known/ directory, a working MCP endpoint, an A2A agent card, a clean crawler path. Miss those and an agent cannot transact with you, even if your site looks perfect to a human.

This checker runs real protocol probes - not LLM guesses. It fetches your .well-known/* files, runs a real MCP initialize handshake, validates A2A agent cards, audits how 20 AI crawlers see your homepage, checks JS parity, and emits copy-paste starter files for every gap.

What agent standards this checker validates

Agent readiness is not one thing - it is a stack of small, machine-readable contracts. We probe each one for real:

• .well-known/openapi.json - the OpenAPI schema that lets an agent call your API.
• .well-known/mcp.json plus a live MCP initialize handshake - the Model Context Protocol endpoint agents use to discover your tools. We do not just check the file exists; we speak the protocol.
• .well-known/ai-plugin.json - the OpenAI plugin manifest.
• A2A agent card - validated against the agent-to-agent schema, not just fetched.
• .well-known/security.txt - the security contact agents and researchers look for.
• Wikidata anchor - whether your brand entity is anchored in the knowledge graph AI engines trust.
• WebMCP markers - inline tags that expose tools without a separate endpoint.
• AI crawler audit - 20 real AI user-agents fetching your homepage, with status and headers.
• JS parity - whether your no-JavaScript HTML still exposes your brand and content.

Why .well-known files matter for AI agents

The .well-known/ directory is the front door for machines. It is where an agent looks first to learn what it can do on your site: which APIs exist, what tools you expose over MCP, who to contact. A human never sees these files. An agent cannot work without them.

Ship them and you become transactable - an agent can read your catalogue, call your endpoints, complete a task. Skip them and the agent falls back to scraping your HTML and guessing, which is slow, lossy, and usually ends with the agent picking a competitor that made itself easy.

How to add an MCP endpoint and A2A agent card

1. Stand up an MCP server. Expose your core actions (search products, check stock, start a return) as MCP tools, then advertise it at /.well-known/mcp.json. The checker's fix-it files give you a working starter manifest.
2. Publish an A2A agent card. Drop a valid agent card at /.well-known/agent.json describing your agent's skills and endpoints so other agents can negotiate with yours.
3. Verify the handshake. A file that 200s is not enough - the MCP initialize call has to succeed. Re-run this checker until the handshake verifies, not just the fetch.
4. Anchor your brand. Add sameAs links to your real Wikidata and social profiles so agents can confirm the entity behind the endpoint is you.

What AI crawlers see without JavaScript

Many storefronts render the important content with client-side JavaScript. Most AI crawlers take the raw HTML and do not execute your JS. If your brand name, products and key copy only appear after hydration, the crawler sees an empty shell. The JS-parity check tells you exactly what an agent reads before a single script runs - and it is often a lot less than you think.

Frequently asked questions

Is this agent readiness checker free?

Yes. Enter a URL and get your readiness score and per-protocol verdicts with no login. The fix-it files and the full 20-crawler audit are one free sign-up away.

Are these real protocol probes or AI guesses?

Real probes. We fetch the actual files, run a real MCP initialize handshake, and validate A2A cards against their schema. No model decides a verdict.

Do I need every standard to pass?

No. Start with the ones that move the needle for your business - usually an MCP endpoint and a clean crawler path - then add the rest. The verdicts are ordered so you fix the highest-impact gap first.

How often can I run it?

Up to 10 checks a day per visitor on the free page. Sign up free to lift the cap, pull the fix-it files, and re-check over time.

Related free tools: AI Crawler Access Checker · Structured Data Tester · Brand Check.