What is available now
WebMCP is a proposed, experimental browser standard for exposing clear website actions to compatible AI clients. Chrome documents two progressive-enhancement approaches: declarative attributes on existing forms and an imperative JavaScript API for dynamic tools. Edge also lists WebMCP as an origin-trial feature, not a universal stable web capability.
The ordinary page must remain complete when WebMCP is unavailable. A tool should describe one narrow action, validate inputs in application code, expose only the minimum result, and preserve visible user control.
Security is part of the interface. Website text and tool output can be untrusted. Tool schemas should be strict, cross-origin tools should not be exposed by default, and actions that spend money, publish, change an account or create an obligation should not run silently. Visibility Score exposes only preparation and explanation, never checkout or payment.
Declarative HTML attributes
The simplest path to WebMCP support is adding attributes to your existing forms. No JavaScript required.
<form toolname="book_appointment"
tooldescription="Book an appointment with our team">
<input name="date" type="date"
toolparamdescription="Your preferred appointment date" />
<input name="service" type="text"
toolparamdescription="Type of service requested" />
<button type="submit">Book Now</button>
</form>
Note: attribute names follow the current W3C explainer (toolname, tooldescription, toolparamdescription with no hyphens). These names may change before the specification reaches a stable release.
Why it matters for your visibility scores
Until now, website visibility has been a three-dimensional problem. Each dimension represents a different way machines discover and use your content.
Agent readiness is the fourth technical dimension in the Visibility Score scanner. It checks whether a site exposes clear, safe and user-visible interactions. That does not prove that a live agent currently uses the site, and consequential actions such as payment still require direct user control.
Robots directives guide crawling, Schema.org describes content, and WebMCP can describe a page action to a compatible agent. They solve different problems and none proves that a live AI service recommends the business.
What to check on your site
These are the practical foundations to audit now.
- Semantic HTML — use real headings, labels, buttons and forms so people and assistive technology can understand the page without an experimental API.
- Narrow tool descriptions — describe one action precisely, distinguish preparation from execution, and use strict input schemas.
- Application validation — treat agent inputs and page content as untrusted and enforce the same server-side checks as any human interaction.
- Human control — keep payment, publishing, account changes and other consequential actions outside automatic tools unless a separately governed confirmation boundary exists.
The bottom line
Agent-friendly structure may make some future interactions simpler, but the standard is still experimental. Build a strong human website first, enhance it safely, and verify real use separately from technical readiness.
Visibility Score currently uses WebMCP only for preparing a readiness scan and explaining an already visible score. It does not expose checkout, payment, contact submission or account changes.