Has your website exposed tools for AI agents?
WebMCP is an experimental proposed web standard that lets a site expose structured, callable tools to supporting browser agents. This checker looks for a direct WebMCP implementation — declarative tool-marked forms, imperative tool registration in script, and an active origin-trial token — plus a few supporting signals that tend to travel alongside agent-ready sites.
Source-level public-page check. No login or form submission is performed.
What this checker looks for
Declarative tools. WebMCP lets a site make an existing HTML form agent-callable with toolname and tooldescription attributes. A form only counts as a registered declarative tool when both attributes are present and non-empty. We scan the fetched page markup for them.
Imperative tools. Sites can also register tools in JavaScript through navigator.modelContext.registerTool(). We statically inspect inline scripts and a small sample of same-origin script files. Tools may be registered dynamically, lazy-loaded, gated behind authentication, or otherwise missed by a static scan — so an absence here isn't conclusive.
Origin-trial token. WebMCP is currently available through a Chrome origin trial. We check the response headers and meta tags for a token. A missing token means the trial isn't enabled for this response — it isn't proof that no WebMCP code exists on the site at all.
Supporting foundations. Separately from WebMCP itself, we report a few signals that tend to help agents and other machines work with a page: an llms.txt summary file, semantic HTML landmarks, ARIA signals, and Schema.org structured data. These are supporting context, not WebMCP implementation, and don't factor into the readiness verdict above.
This checks one signal. The full audit checks 32.
Run Free Audit