WebMCP / readiness
WebMCP readiness, honestly
"Is my site WebMCP ready?" is a fair question with an uncomfortable answer: partly measurable, and probably not urgent. Here is what readiness actually means, what any check can and cannot see, and how to decide whether to act.
What "readiness" means
To be WebMCP ready is to expose your site's key actions as tools an in-browser agent can call, described clearly enough that an agent picks the right one and passes the right arguments. Not every action, and not for its own sake: the high-value ones a person would ask an assistant to do on your site, search, book, quote, add to cart, contact. Readiness is a design decision about which capabilities to publish and how to describe them, not a checkbox.
What a check can and cannot see
Declarative tools are HTML attributes on a form, so they sit in the page source and a static scan can see them. Imperative tools are registered at runtime by JavaScript, so they do not exist in the source at all; only a browser that executes the page would observe them. This is not a flaw in any particular scanner. It is a property of the standard. The honest consequence: a report saying "no WebMCP detected" may be describing a site that registers tools perfectly well the moment it runs. Detection can confirm presence; it cannot reliably confirm absence.
A missing tool is not a defect today
Because no mainstream agent consumes WebMCP tools yet, there is no ranking effect, no penalty, and no user who is currently underserved by their absence. So the absence of WebMCP is not, right now, a website defect. This is where the field's incentives get dangerous. A "WebMCP score", or an alert that your missing document.modelContext is a problem to fix immediately, manufactures urgency the facts do not support. AIOFacts will not publish such a score, and you should be wary of anyone who does. What can be reported honestly is factual and narrow: does this page's HTML contain declarative tool annotations, and does the page reference the API at all? That is evidence a person can verify, not a grade.
Seeing the evidence on a specific site
AIOFacts defines the standard. To look at what a particular page publishes, AIOInsights runs an early-preview WebMCP readiness check. It is labelled beta on purpose: it reports the observable evidence a site exposes, with the same caveat above, that a static view cannot see runtime registration. It does not hand you a number and call it a verdict. That division is deliberate: the reference and the diagnostic are different jobs, and keeping the diagnostic on AIOInsights is what lets this page stay a reference.
Deciding whether to act
A reasonable posture for most businesses, as of mid-2026:
- Understand it now. Knowing what WebMCP is, and where the agentic web is heading, costs little and positions you to move fast when a real consumer ships.
- Act where it is cheap and reversible. If you already have a clean search or booking form, adding declarative annotations is low-cost and easy to remove. That is a defensible early move.
- Do not rebuild around it yet. Betting a roadmap on a draft that changed its own namespace this month, with no agent consuming it, is premature. The API is moving and the payoff is not here.
- Watch the trigger. The moment to invest is when an agent people actually use, Gemini in Chrome is the announced first, begins calling these tools. That is a fact you can wait for.
None of that is investment advice about your specific site; it is a description of where the standard stands. What to do about your situation depends on your business, and that judgement is what Digilu works through with you rather than selling you an implementation of a feature nothing consumes yet.
Sources
- webmachinelearning/webmcp: the specification and its two APIs.
- AIOInsights WebMCP readiness beta: the first-party check this page refers to.
Reviewed 23 July 2026. The "no mainstream consumer" status is the single fact most likely to change; this page will be updated and the change dated when it does.
Zoom out
WebMCP is the action layer. There is a picture around it.
Before an agent acts on your site, it has to understand you. That is where structured data, clear entities, and readable content come in.