What WebMCP is
WebMCP is Google's proposed open web standard in which sites expose their own "tools" — JavaScript functions, HTML forms — in a machine-readable way, so that a browser-based AI agent can trigger them with precision instead of mimicking clicks on an interface designed for human eyes.
The idea is simple but the implications run deep. Today, an agent booking a flight has to visually recognize an on-screen calendar and click the right cell. With WebMCP, the site explicitly declares here is the selectDate function, with these parameters, and the agent calls it. Faster, more reliable, less fragile when the front-end gets redesigned.
The origin trial ships in Chrome 149
The experimental origin trial is live as of Chrome 149, and Gemini in Chrome will soon support WebMCP APIs — turning the browser into a first-class actor in the so-called agentic web. Alongside it Google shipped **Chrome DevTools for agents**, already supported by Antigravity and more than 20 other coding agents, to give developers proper debugging surfaces for the new paradigm.
Who's already on board
Per eWeek and SD Times, Google named Booking.com, Expedia, Instacart, Intuit, Shopify and Redfin among the brands already experimenting with WebMCP to design agent-friendly experiences. It's a clear sign that Mountain View is trying to make WebMCP a de facto standard before W3C groups formalize it — much as it did with AMP and Web Push.
Why it matters
For people working on the web, WebMCP reframes classic SEO: optimizing for Google's crawler is no longer enough. You also need to design an agent-readable API. It's a new optimization surface that, if it catches on, will become mandatory for anyone who wants to stay visible in an internet whose visitors aren't only human.