How do I make my website agent-ready with WebMCP?

Make the basics agent-ready first: pages served as HTML, clear facts, structured data and forms that work without tricks. WebMCP, which lets a page expose tools such as "book a slot" to AI agents, is still an early preview: Chrome opened it in February 2026 and the October 2026 draft is not a W3C standard. Experiment with it, but do not rebuild for it yet.

Facts

Chrome announced WebMCP as an early preview for program participants on February 10, 2026. (Chrome for Developers: WebMCP early preview (February 2026))
The WebMCP specification dated 9 October 2026 is a Draft Community Group Report and states it is not a W3C Standard nor on the W3C Standards Track. (W3C Web Machine Learning Community Group: WebMCP draft (October 2026))
Chrome describes two WebMCP APIs: a declarative one for actions in HTML forms and an imperative one in JavaScript. (Chrome for Developers: WebMCP early preview (February 2026))
WebMCP borrows its idea of tools from the Model Context Protocol, which Anthropic introduced as an open standard on November 25, 2024. (Anthropic: Introducing the Model Context Protocol)
Google states that structured data helps it understand the content of a page; its documentation lists the supported types and their rules. (Google Search Central: introduction to structured data)

What is WebMCP?

WebMCP is a proposed browser API that lets a web page offer its own functions to an AI agent as tools, each with a name, a description and an input schema. Instead of guessing which button to click, the agent calls "check availability" or "add to cart" directly.

The current draft puts the entry point at document.modelContext; earlier write-ups used navigator.modelContext, a sign of how fast it is still moving.

// WebMCP draft (W3C Community Group, 9 October 2026): the API may still change.
if ("modelContext" in document) {
  document.modelContext.registerTool({
    name: "check_availability",
    description: "Returns the free appointment slots for a date (YYYY-MM-DD).",
    inputSchema: { type: "object", properties: { date: { type: "string" } }, required: ["date"] },
    execute: async ({ date }) => (await fetch("/api/slots?date=" + date)).json(),
  });
}

Should my business implement WebMCP now?

Only as an experiment, if your site has actions an agent would really perform (booking, quoting, checking stock). The draft can change, and Chrome itself calls it an early preview.

For most company websites the work that pays now is the one agents and AI search already use: readable pages and correct facts.

What makes a site agent-ready today?

The same things that let AI assistants read and cite a page:

Layer	What to do	Status in 2026
Readable content	Serve the main text as HTML, not only after JavaScript runs	Works today
Clear facts	Prices, hours, areas and policies written on the page	Works today
Structured data	Organization, Product, LocalBusiness, FAQ in JSON-LD	Works today
Usable forms	Labelled fields, no hidden steps, clear confirmations	Works today
WebMCP tools	Expose actions as tools with a schema	Early preview

Who should build it?

The developer or coding agent that already works in your repository: these are code changes, not settings. With easyIndex, that agent builds the readable pages and structured data over MCP, and easyIndex audits and checks each page on the live site.

Step by step

Check what agents can read. Open the page source: the main text should be in the HTML.
Write the facts down. Prices, hours, service areas and policies on the pages, not only in images or PDFs.
Add structured data. JSON-LD that matches what the page says.
Clean up the forms. Labelled fields and a clear confirmation an agent can read.
Prototype WebMCP on one action. Behind a feature check, in a branch, following the current draft.

Common mistakes

Rebuilding the site for a draft API before the basics are readable.
Exposing a tool that changes data (an order, a payment) without confirmation from the person.
Copying old tutorials that use an entry point the current draft has renamed.
Hiding prices and terms behind JavaScript or forms that an agent cannot read.

How easyIndex solves it

easyIndex gives your coding agent one step at a time over MCP to make the pages readable, factual and marked up, then audits each page, checks it live and measures AI visitors and AI crawler reads, and records which sources ChatGPT, Claude, Perplexity and Gemini cite for your tracked questions in the monthly citation scan your agent runs. It costs US$ 30 per website a month, with a 7-day free trial.

Frequently asked questions

Is WebMCP the same as MCP?

No. MCP connects an agent to servers; WebMCP lets a web page in the browser offer tools to an agent. WebMCP borrows MCP's idea of tools.

Which browsers support WebMCP?

Chrome opened it as an early preview in February 2026. Check the current draft before building on it.

Does WebMCP help me appear in ChatGPT answers?

Nothing shows that today. Answers come from pages the assistant can read; WebMCP is about actions.

Is llms.txt part of being agent-ready?

It is a separate, optional file. Readable pages and structured data matter more.

Continue reading

Last verified on October 9, 2026.