Inblog
Start your company blog
GEO

WebMCP

WebMCP (Web Model Context Protocol) is a proposed W3C web standard that lets websites expose what they can do as structured tools that in-browser AI agents can discover and call directly. Instead of agents screenshotting pages and guessing where to click, the site declares its capabilities—inverting control back to the website.

WebMCP (Web Model Context Protocol) is a proposed W3C web standard that lets websites expose what they can do as structured tools that in-browser AI agents can discover and call directly. Instead of agents screenshotting pages and guessing where to click, the site declares its capabilities—inverting control back to the website.

Why It Matters

As AI agents increasingly operate websites on behalf of humans, reverse-engineering human-oriented HTML has proven slow and error-prone. WebMCP was co-authored by Google and Microsoft engineers in the W3C Web Machine Learning Community Group: an initial explainer appeared in August 2025, followed by a Draft Community Group Report on February 10, 2026—the same day Chrome 146 Canary shipped a flag-gated preview. As of June 2026 it has progressed to a Chrome origin trial that real sites can test. Lighthouse 13.3, released in May 2026, added an "Agentic Browsing" audit category that checks for WebMCP integration by default. Google's John Mueller weighed in too: in June 2026 he called llms.txt "purely speculative for now" while saying he likes the WebMCP approach for its clear goals and processes.

How It Works

A web page registers tools through a JavaScript API (registerTool on modelContext). Each tool carries a name, a natural-language description, a JSON Schema input definition, and an execute callback. A site might declare tools like "search flights," "add to cart," or "request a quote," and in-browser agents discover and invoke them using the same tool use pattern LLMs already follow. Dozens of screenshot-interpret-click cycles collapse into a single structured tool call, improving both accuracy and speed. Because tools can wrap existing client-side JavaScript, no page re-architecture is required, and a declarative variant using annotated HTML forms is also under discussion.

Relationship to MCP and llms.txt

  • Model Context Protocol: A server-side standard where you run a separate MCP server to offer tools to AI clients. WebMCP brings the same philosophy to the web page itself (client side), letting agents work within the user's logged-in browser session and your existing UI logic.
  • llms.txt: A static text file that describes site content—it helps agents read. WebMCP is a callable interface that helps agents act. The two are complementary, not competing.

No browser besides Chrome has announced an implementation timeline yet, though Microsoft's co-authorship of the spec makes Edge support likely.

Sources:

How inblog Helps

Registering WebMCP tools takes engineering work, but the prerequisite—agent-readable content—is something you can build today. inblog publishes posts with semantic HTML and automatic structured data (JSON-LD), making blog content easy for agents to parse. Organizing your product docs, FAQs, and guides on inblog lays the content foundation that your site's tools will work alongside once agentic browsing goes mainstream.