All insights
AI Product Building AI Agents Architecture

WebMCP turns websites into agent-native interfaces

Chrome's MCP integration lets websites expose structured tools to agents instead of agents scraping and guessing at UI elements

Chrome for Developers — WebMCP announcement (https://developer.chrome.com/blog/web-mcp) · · 5 connections

Today’s browser agents work by screenshotting pages, reading HTML, and guessing which elements are buttons — essentially screen-scraping with extra steps. WebMCP flips this: websites expose structured tools directly via navigator.modelContext.registerTool(), telling agents exactly what actions are available and what parameters they accept. Instead of guessing “this looks like a search box,” the agent knows “call searchFlights(from, to, date).”

This is the practical realization of The UI moat collapses — API quality becomes the purchasing criterion — when every website can expose its capabilities as structured tools, the visual interface becomes irrelevant for agent interactions. It accelerates B2B becomes B2A — agents become the buyer because agents can now evaluate and transact with services programmatically. The pattern mirrors In agent-native architecture, features are prompts — not code: the website’s value shifts from its rendered interface to the structured capabilities it exposes. WebMCP is to websites what MCP was to developer tools — a standard protocol that makes the entire web ai-agents-traversable. Each registerTool() call is a formal contract, and Tools are a new kind of software — contracts between deterministic systems and non-deterministic agents explains why these interfaces must be designed for non-deterministic consumers who rely on descriptions, not source code.