What is WebMCP, and should your website have it?
WebMCP lets a website describe tools that an AI agent can use while working in a browser. A tool might search a catalog, choose an option, or submit an inquiry. The page provides a defined action and input structure for the agent.
For a business website, its value is practical: an agent can work with the site’s intended actions while the customer uses a browser-based experience.
How a tool differs from a visible control
An ordinary website presents links, fields, and buttons. A browser agent can try to interpret those controls and interact with them. A WebMCP tool supplies a description of the action and the data it expects.
For an inquiry, that description can identify the request’s purpose, required fields, and available choices. The result can tell the agent whether the submission succeeded or what needs correction.
The website still needs a useful human interface. People may complete the same task themselves, review the agent’s work, or take over partway through it.
How it differs from remote MCP
Remote MCP connects an AI application to a server endpoint. WebMCP works within the browser page and its context. A browser must visit the website to discover its page tools; publishing WebMCP tools does not by itself place a website into every AI app’s list of connections.
Consider where your customer’s agent will operate. A browser task may benefit from WebMCP. A chat app connecting directly to a site’s tools may use remote MCP. A website can provide both interfaces to the same underlying actions.
Where it can help
Tools are useful when a task has a clear operation behind it. On a service-business website, that could be submitting a structured estimate request. On another site, it might involve a search or a choice within an existing workflow.
Begin with one action customers already want to complete. Write a precise tool description, define the inputs, and make the result understandable. Confirm that the same business rules apply whether a person or an agent performs the action.
Do not give an action a broader meaning than it has. An inquiry confirmation should say that the request was received; it should not imply that availability or pricing has already been confirmed.
Check the actual browser support
WebMCP support is evolving. Chrome’s documentation describes an origin trial and a local testing flag. Check the current browser and agent combination you intend to support, and test it directly. Chrome’s WebMCP documentation.
Use the browser’s available inspection tools to confirm which tools the page registers and what inputs they accept. Test a successful request, a missing required field, and an unavailable action. A failure should leave the agent with useful information about the next step.
Should your site have it?
WebMCP is worth evaluating when your visitors use browser agents and your website has useful actions those agents should complete. First make the underlying action reliable and the human experience clear. Then expose the action in a way the agent can understand.
Wirechunk includes WebMCP tools for its published-site inquiry workflow. See how agent-ready websites work.