The Three WebMCP Lighthouse Audits (and How to Pass Them)

What Are the WebMCP Lighthouse Audits?

The WebMCP audits are the part of Lighthouse's agentic browsing category that check whether your site exposes its tools to AI agents through the navigator.modelContext API. There are three of them, and they only apply once you have adopted WebMCP — until then, Lighthouse marks them Not Applicable and they do not affect your pass-ratio^[1].

If you want the full picture of the category they live in, start with the Lighthouse Agentic Browsing guide.

How Do You Pass the WebMCP Audits?

Three steps clear all three audits:

• Register your tools. Call navigator.modelContext.register() for each capability you want agents to use, or declare tools in HTML with toolname and tooldescription.
• Cover your key forms. Add declarative WebMCP attributes to high-intent forms — quote requests, search, sign-up.
• Validate every schema. Give each input a name, every parameter a toolparamdescription, and keep tool attributes symmetric. This is the only one of the three that can fail outright.

For the standard itself, the API, and what to ship first, see the deep dive: WebMCP: The W3C Standard That Makes Your Website Agent-Ready.

Frequently Asked Questions

Do the WebMCP audits lower my score if I have not adopted WebMCP?

No. Lighthouse marks all three as Not Applicable when no WebMCP is detected, so they are excluded from your pass-ratio. You can still score 3/3 on the applicable stability and accessibility audits.

Which WebMCP audit can actually fail?

Only webmcp-schema-validity emits a failure. It fails on asymmetric tool attributes, inputs without a name, and missing toolparamdescription. The other two are informational.