Documentation Run a scan

3 tools · no account · no key

Use all of this from your own agent

Everything on this site is available to an agent, not just to a person. That is the same standard we hold other sites to, so it would be a poor look to publish a scanner that only a human can drive. Connect it to your AI client and it can scan a site, read the reference, build the file and hand you the publishing steps without anyone opening this page.

  1. Connect it

    One Streamable HTTP endpoint, no account and no key. Add it to Claude, or to any client that speaks MCP.

    https://agentic-readiness.lumar.io/mcp
  2. Then ask for the whole job, not a step of it

    Three tools arrive with it — scan_site, search_docs, build_agent_file — and they compose into the work: find what is broken, read why it matters, generate the fix, check the fix against the same rules the scan uses.

    "Scan example.com, then build whatever agent files it is missing and tell me how to publish them on Cloudflare."
  3. It interviews you rather than guessing

    Asked to build something, it first reports what the site already publishes and what still needs deciding, in words your agent can put to you directly. Only then does it generate. An artifact nobody agreed to is not a feature, so the questions are deliberate and the answers are yours.

  4. It refuses to hand over its own defects

    Every generated file is run through the analyzer this scanner scores live sites with, and anything that fails is withheld with the reason. If a card would advertise an endpoint that does not answer, it will not write the address until the endpoint answers.

  5. The rest of the surface, for whatever your agent speaks

    In a browser that supports it, each builder also registers its tools in the page itself, so a browsing agent can complete the wizard while you watch.