A new W3C proposal, supported by Google and Microsoft, aims to enable developers to expose client-side JavaScript features to AI agents, fostering cooperative interactions between human users and AI within the same web environment.
Members of the World Wide Web Consortium (W3C), including tech giants Google and Microsoft, have unveiled the WebMCP API. This new JavaScript interface is designed to allow web applications to expose client-side “tools” to artificial intelligence agents. The intention is for this API to enable AI agents to interact directly with web pages and engage in joint operations with human users within the same online environment.
The WebMCP API is currently accessible for early access via Google, as stated by Andre Cipriani Bandarra, a developer relations engineer for Chrome and the web at Google, in a blog post published on February 10. Bandarra explained that “WebMCP aims to provide a standard way for exposing structured tools, ensuring AI agents can perform actions on your site with increased speed, reliability, and precision.”
A preliminary community group report concerning WebMCP was released on February 12 by the W3C Web Machine Learning Community Group. Within this report, the WebMCP API is characterized as a JavaScript interface allowing web developers to present web application functionalities as “tools.” These tools are defined as JavaScript functions accompanied by natural language descriptions and structured schemas, capable of being invoked by agents, browser-based agents, and assistive technologies. The report indicates that web pages leveraging WebMCP can function as Model Context Protocol (MCP) servers that implement tools through client-side scripting rather than server-side, thereby facilitating cooperative workflows where both users and agents collaborate within the same web interface. The report’s editors include Khusal Sagar and Dominic Farolino from Google, along with Brandon Walderman from Microsoft. It’s important to note that the specification is neither a formal W3C standard nor currently part of the W3C Standards Track.
Bandarra highlighted several potential applications, such as customer assistance, e-commerce, and travel services, where AI agents could help users complete support forms, find products, or make flight reservations. He detailed two proposed APIs within WebMCP designed to empower browser agents to act on a user’s behalf: a declarative API for standard actions definable directly within HTML forms, and an imperative API for more intricate and dynamic interactions requiring JavaScript execution. Bandarra commented, “These APIs serve as a bridge, making your website ‘agent-ready’ and enabling more reliable and performant agent workflows compared to raw DOM actuation.”
