WebMCP
WebMCP is a draft web standard that lets a website offer AI agents ready-made actions, like search the catalogue or check delivery, in the browser.
WebMCP is a draft web standard that lets a website offer AI agents a set of actions they can use directly in the browser. A shop might offer "search products," "add to basket" and "check delivery to a postcode." An agent visiting the page can call those actions instead of finding its way around by clicking buttons and reading the screen.
It's published as a draft by a W3C community group, which says plainly that it isn't a W3C standard and isn't on the standards track. It's early. But it shows where things are heading for brands whose customers will send agents to shop for them.
How it works
The name comes from MCP, the Model Context Protocol, a common way for AI systems to call tools. WebMCP brings the same idea to web pages. The page registers its tools in JavaScript, each with a name, a plain-language description and a description of the input it needs. An agent in the browser can list those tools and run them.
The tools run in the page itself, so they use the same session as the person, with the same basket and the same sign-in.
Why it matters
Agents are getting better at using websites the way people do, but it's slow and they make mistakes. A page that offers clear actions is easier for an agent to use, and so more likely to be the shop it completes the job on.
An example
For example, Harrow & Pine adds a WebMCP tool called "check delivery date," which takes a postcode and a product and returns the earliest delivery day. Someone asks their AI browser to find an oak table that can arrive before Christmas. On Harrow & Pine's site, the agent calls the tool and gets an answer in one step. On a competitor's site, it has to open each product, find the delivery section and read it.
What to do now
Watch it rather than build on it. In the meantime, the basics already help agents: clear HTML, forms with proper labels, and product details in the page's text.