Skip to content

MCP server

For agents, not for people

Rheono runs as a remote MCP server. Your agent calls it before a deploy, a handover or a client delivery, and gets back what is wrong on the live host in plain words. Nothing is installed locally and there is no SDK to wire up.

Endpoint

https://rheono.dev/api/mcp

Transport: streamable-http. No API key.

Connect

  • Cursor

    Settings → MCP → Add new MCP server, transport `streamable-http`, URL above. Or install the plugin from the marketplace.

  • Claude

    Settings → Connectors → Add custom connector, paste the URL. Works on any paid plan.

  • ChatGPT

    Settings → Connectors → Advanced → Developer mode, then add the URL as a custom connector.

  • Grok

    Add the plugin from the Grok Build marketplace, or register the URL as a remote MCP tool in the Grok API.

Tools

  • check_app

    Read-only pass on a live URL: what the TLS handshake gives back, missing security headers, cookie flags, version disclosure and conventionally exposed paths.

  • verify_ownership

    Issues and checks a DNS TXT record that proves a domain is yours, so later runs can reference it as an owned asset.

  • request_pentest

    Hands the domain to a researcher for the deeper read-only probe. The report is emailed.

  • check_status

    Reports whether a request exists for a domain and when it was last checked.

What a run touches

A check is plain web requests to the host you name: a TLS handshake, a request for the page, a handful of conventional paths, and the header data that comes back. No logins, no payloads, no port scan. Private, loopback and cloud-metadata addresses are refused, so the server cannot be turned into a probe of your own network.

Rate limits are per client and per tool, so a looping agent cannot drain the endpoint. The same scan policy and opt-out apply whether the check was started by an agent or by a person.

Verification and the report

A shallow check needs nothing proven. To have findings tied to you, publish the TXT record verify_ownership returns at _rheono.<domain>; the token is public, so it is safe in DNS. Unverified domains still get a report, just a shorter one.

The deeper probe is still read-only and runs at about one request per second. When it finishes you get the report by email, and a researcher can start a full engagement from there if you want one.

Limits

The server is anonymous and read-only. It holds no account state, so a domain is only “yours” in the sense that its TXT record says so. Findings are public posture, not your code or data. Start at the free check if you want to see the shape of a report first.

Questions or a broken connection: sales@rheono.dev.