WebMCP: make your website actionable by AI agents
WebMCP lets your page expose tools that AI agents call directly — instead of guessing where to click. Animam consumes it today via a polyfill, in every browser.
What is WebMCP?
WebMCP (Web Model Context Protocol) is a W3C Web Machine Learning Community Group standard. It lets a web page publish "tool contracts" that AI agents call directly, through the browser API navigator.modelContext. It replaces the "screenshot → vision model → guess the click" pattern with reliable function calls.
The catch: today only Chrome 146+ (behind a flag) consumes WebMCP natively. If you implement it, almost no agent can use it yet.
The Animam take: forward-compatible
The Animam widget polyfills navigator.modelContext. You write standard WebMCP once, and it works immediately with our agent — without waiting for the ecosystem.
Standard, not proprietary
You write standard WebMCP (navigator.modelContext). No house API to learn or maintain.
Works everywhere, today
Our polyfill makes your tools usable in every browser, without waiting for Chrome to ship the API.
Ready for native agents
The same code will serve Chrome’s native agent (and Claude-in-Chrome) when it ships. You’re additive.
Secure by design
The agent only sees the schema; mutating actions require confirmation; your protections apply.
Make your site actionable
navigator.modelContext.registerTool({
name: 'add_to_cart',
description: 'Add a product to the cart.',
inputSchema: {
type: 'object',
properties: { sku: { type: 'string' }, qty: { type: 'integer' } },
required: ['sku'],
},
execute: async ({ sku, qty }) => {
const r = await fetch('/api/cart', {
method: 'POST', headers: { 'Content-Type': 'application/json' },
body: JSON.stringify({ sku, qty }),
})
return await r.json()
},
})Details and examples in the guide.
FAQ
What is WebMCP?
WebMCP (Web Model Context Protocol) is a W3C Web Machine Learning Community Group standard. It lets a web page expose structured tools to an AI agent through the navigator.modelContext browser API. Instead of screenshotting and guessing where to click, the agent calls functions the page declares directly.
Is WebMCP the same as MCP?
No. MCP (Model Context Protocol) connects a server to a client, on the backend. WebMCP connects a page to an in-page agent, in the browser (navigator.modelContext). Same tool-contract idea, two distinct and complementary specs. Neither is tied to a specific model (Claude, Gemini, GPT).
Which browsers support WebMCP?
Natively, only Chrome 146+ behind a flag (early preview). But the Animam widget polyfills navigator.modelContext, so your standard WebMCP tools work with the Animam agent in every browser today, and with Chrome’s native agent once it ships.
Do I need to code for each AI agent?
No. You write standard WebMCP once. The same registration serves the Animam agent and future browser-native agents — you’re additive, with no proprietary API to maintain.
Is it secure to let an agent act on my site?
Yes. The agent only receives the tool schema, never your code or secrets. State-changing actions (cart, booking, payment) require explicit visitor confirmation. Tools run with the visitor’s session, so your auth, CSRF and rate-limiting apply.
How is this different from a chatbot?
A chatbot answers. With WebMCP, the agent acts: it adds to cart, books a slot, filters a catalog, starts a flow — directly, instead of explaining how to do it.