Google's WebMCP turns websites into structured data sources for AI agents

Google has introduced WebMCP (Web Model Context Protocol), a new JavaScript interface designed to enable AI agents to communicate with websites in a standardized way. As Google developer André Cipriani Bandarra explains in a blog post, WebMCP transfers the idea of the Model Context Protocol to websites, allowing AI agents to interact with web pages in a structured manner within the browser – without having to scrape the page's DOM and navigate through the code.

The interface allows AI agents to perform specific actions on websites, such as filling out forms, searching for and ordering products, or booking flights. "WebMCP aims to provide a standard for structured tools to ensure that AI agents can perform actions with increased speed, reliability, and precision," says Bandarra.

WebMCP extends the Navigator interface with navigator.modelContext, featuring methods like registerTool() and provideContext(). Google distinguishes between a declarative API for simple form actions and an imperative API for complex JavaScript interactions. Each registered tool requires a name, a natural language description, a JSON schema for inputs, and an execute callback function.

Differences to Microsoft's NLWeb

In mid-2025, Microsoft introduced NLWeb, its own open-source project for the "agentic web." The key difference: NLWeb is designed server-side, turning websites into MCP servers with natural language interfaces. WebMCP, on the other hand, runs in the browser within the JavaScript context, without separate backend components.

According to the official WebMCP specification from the W3C Web Machine Learning Community Group, Google and Microsoft are now collaborating on the WebMCP standard. Alongside Google developers, Microsoft developers are also listed as co-authors, suggesting a potential convergence.

WebMCP is currently available as an Early Preview through Google's Early Access Program for developers and is available in Chrome 146 as a DevTrial. Google may integrate the interface into Chrome and its AI services like Gemini in the future to allow browser agents to interact directly with websites.

(odi)