WebMCP: Chrome's New Standard for AI Agent-Ready Websites
WebMCP is a proposed browser standard by Google that lets website owners define structured tools and actions for AI agents directly in the browser, replacing fragile DOM scraping with reliable, typed interfaces. If you've been watching the AI agent space, you know the problem: agents like ChatGPT, Claude, and OpenClaw try to interact with websites by clicking, scraping, and navigating the DOM — essentially imitating what a human would do. It works, but it's fragile, slow, and often breaks.
With ChatGPT now serving 900 million weekly active users (OpenAI, 2025) and Google AI Overviews reaching 1.5 billion users monthly across 200+ countries, the volume of agent-driven web interactions is growing fast. AI-referred website sessions grew 527% between January and May 2025 (SparkToro), and the trend is accelerating.
Google just shipped a better answer. On February 10, 2026, Chrome's DevRel team announced WebMCP — and it's one of the most significant shifts in how the web will work in the agent era.
What Is WebMCP?
WebMCP is a proposed browser standard that lets website owners define structured tools for AI agents — similar to how MCP (Model Context Protocol) works in desktop applications, but natively in the browser.
The goal: instead of an agent blindly clicking through your checkout flow, your website tells the agent exactly what actions are available, what data they need, and how to call them. Fewer errors. Faster execution. No DOM scraping required.
WebMCP proposes two APIs:
Declarative API — for standard actions that can be defined directly in HTML forms. Think of it as semantic HTML for agents: you describe what a form does, what fields it needs, and what it returns.
Imperative API — for complex, dynamic interactions that require JavaScript. Flight booking with live seat maps, dynamic pricing, multi-step flows — anything that doesn't fit a simple form.
Why Does WebMCP Matter for Agent Readiness?
AgentSpeed already checks whether your website is ready for AI agents. WebMCP adds a new dimension: not just can agents access your site, but how well can they interact with it?
Right now, most agents fall back to DOM navigation — they read your page like a human would, interpret the structure, and try to click the right buttons. This is why cookie walls, CAPTCHA, and JavaScript-heavy UIs block them so effectively.
With WebMCP, agents that support it (and Chrome-based agents will) can skip the DOM entirely and call your defined tools directly. A flight booking agent doesn't need to render your page — it just calls searchFlights(origin, destination, date) and gets structured results back.
What Are the Key Use Cases for WebMCP?
Google's announcement focuses on three verticals that illustrate the power of WebMCP:
Customer Support — Users can create detailed support tickets, with agents automatically filling in technical context. Instead of a user describing their problem in a text box, an agent collects system info, error logs, and context, then submits a complete ticket via the WebMCP tool.
E-Commerce — Agents can search, configure, and purchase products with precision. Your addToCart(productId, quantity, variant) tool gives agents a reliable, structured path to checkout — no scraping required.
Travel & Booking — Agents can search flights, filter results, and book with structured data rather than interpreting your UI. The example is direct: an agent calling searchFlights({from, to, date}) and getting back a typed response with availability and pricing.
How Does WebMCP Relate to MCP?
If you've been following the AI tooling space, MCP (Anthropic's Model Context Protocol) has become the dominant standard for connecting AI models to external tools and data sources. WebMCP is Chrome's take on bringing this pattern to the web itself.
The key difference: MCP runs server-side (your backend exposes tools to an AI). WebMCP runs browser-side (your website exposes tools to a browser-based agent). For users with local AI agents — personal assistants running on their devices — this is huge. The agent never needs to hit your API directly; it interacts through the browser as a trusted proxy.
How Can You Prepare for WebMCP Now?
WebMCP is currently in Early Preview (Early Partner Program access only). But you can start preparing today:
1. Identify your key agent workflows. What would an AI agent most want to do on your site? Book something? Find information? Submit a form? List these explicitly — they'll become your WebMCP tools.
2. Structure your existing HTML semantically. The Declarative API builds on well-structured HTML. Sites with clean semantic markup will adapt faster when WebMCP tooling becomes available.
3. Check your robots.txt and llms.txt. Before worrying about WebMCP, ensure agents can access your site at all. Scan your site on AgentSpeed — if you're blocking GPTBot or ClaudeBot, you have bigger problems than WebMCP readiness.
4. Watch the Early Access Program. Google's EPP gives you access to documentation, demos, and early feedback. Sign up here if you're building a site where agent interaction matters.
5. Design your API for agents, not just humans. Whether or not WebMCP takes off, structured APIs are a better foundation than DOM scraping. REST or GraphQL endpoints with clear semantics will make your site more agent-compatible regardless of which standard wins.
Where Does WebMCP Fit in the Bigger Picture?
WebMCP is part of a broader wave of standards emerging around agentic web interaction. Google's Project Mariner, Anthropic's Claude for Chrome, OpenAI's browser-integrated agents — they're all converging on the same insight: the web needs to be redesigned for agents, not just humans.
The parallel to the mobile era is instructive. When smartphones became mainstream, websites that hadn't adapted their layouts became second-class citizens. We're at the beginning of a similar transition. Sites that define structured agent interfaces — through WebMCP, through well-structured llms.txt, through clean APIs — will have a significant advantage as agent-driven traffic grows.
The question isn't whether this transition will happen. It's whether your site will be ready when it does.
Is your website ready for AI agents today? Run a free scan on AgentSpeed — get your score in seconds and see exactly what's blocking agents from interacting with your site.