Microsoft and Google Propose WebMCP Standard

Engineers from Microsoft and Google have unveiled a new web standard proposal called WebMCP. This allows web pages to directly declare their capabilities to AI agents as tools, enabling agents to interact with web apps through clear interfaces instead of interpreting screenshots or the DOM.

Necessity and Features of WebMCP

Existing Computer Use or browser automation methods are vulnerable to layout changes, incur token costs, and yield uncertain results. WebMCP presents a client-side alternative by defining tools within page scripts to address these issues. This approach leverages live sessions within the browser tab, including user login states and active contexts, and allows the reuse of existing JavaScript logic without the need to build a separate backend server.

Relationship with MCP and Implementation Status

WebMCP is not a replacement for the backend integration protocol MCP (Model Context Protocol) but rather a complement. While MCP connects server-side data and logic, WebMCP serves as a frontend layer that enables agents to immediately interact with the live website UI the user is visiting. Origin trials are currently underway in Chrome 149 and Edge 150, and the Brave browser has also begun experimental support. The ChatGPT desktop app supports this feature as 'Site tools', but WebKit has officially expressed opposition to the proposal, meaning standardization discussions are still ongoing.