Google Unveils WebMCP in Chrome 149 and HTML-in-Canvas API for Agentic Web
Updated
Updated · The New Stack · May 19
Google Unveils WebMCP in Chrome 149 and HTML-in-Canvas API for Agentic Web
7 articles · Updated · The New Stack · May 19
Chrome 149 beta will start an origin trial for WebMCP, an open standard Google says lets websites expose functions and forms directly to AI agents instead of forcing them to navigate pages via screenshots or the DOM.
More than 20 coding agents can now tap Chrome DevTools 1.0 capabilities — including console logs, network traffic and accessibility trees — through Chrome’s built-in MCP server or the DevTools CLI for automated debugging.
Google also launched an early preview of Modern Web Guidance, which teaches coding agents which Baseline web features to use and links with the Google Analytics API so developers can match targets to their users’ browser support.
Separately, Google proposed an HTML-in-Canvas API that would render live DOM elements inside canvas for searchable, accessible and translatable 3D-style interfaces, though Firefox and Safari have not adopted the standard.
With AI agents browsing for us, will the visual design and user experience of websites become entirely obsolete?
As AI agents gain direct web access, how will we prevent them from being weaponized for automated fraud and abuse at scale?
If AI agents bypass ads to complete tasks, how will the internet's digital advertising economy survive this fundamental shift?
2026’s Agent-Ready Web: WebMCP, HTML-in-Canvas, and the Rise of Agent Optimization in Chrome Origin Trials
Overview
WebMCP is a major step forward for the web, making it easier for AI agents to understand and interact with websites. By adding semantic annotations, structured data, and new APIs directly into HTML, CSS, and JavaScript, WebMCP lets websites clearly communicate their structure and purpose to AI agents. This moves agents beyond unreliable visual cues and guesswork, enabling a deeper, machine-readable understanding of web content and user intent. As a result, agent interactions become more reliable and efficient, paving the way for an 'agent-ready' web experience that benefits users, developers, and brands alike.