WebMCP for Retailers: Exposing Catalog APIs directly to Chrome AI Agents

The Model Context Protocol is moving to the browser. WebMCP provides websites a native way to expose catalogs, search APIs, and checkouts directly to Chrome-native AI agents.

For over a decade, digital discovery relied on a human clicking a blue link. But the rollout of Chrome 149's experimental origin trial for WebMCP (Web Model Context Protocol) has changed the game. AI agents can now browse, compare, and transact autonomously.

To support this shift, websites must publish a structured manifest file: webmcp.json. This acts as the machine-readable directory for your business, detailing exactly what operations a bot is permitted to run. Let's look at the implementation.

Creating Your webmcp.json Manifest

The manifest file must reside at your root domain (https://yourdomain.com/webmcp.json). It declares the APIs and functions that agents are allowed to call.

{
  "protocol": "WebMCP/1.0",
  "endpoints": {
    "actions": "/api/webmcp/actions"
  },
  "actions": {
    "search_catalog": {
      "description": "Search product inventories by text",
      "parameters": {
        "query": { "type": "string", "required": true }
      }
    }
  }
}

Declarative vs. Imperative WebMCP

WebMCP supports two integration paradigms:

• Declarative APIs: Annotating standard HTML forms with WebMCP attributes so agents can fill them out natively.
• Imperative APIs: Exposing specific JavaScript functions that the browser-native agent calls directly through the client window context.

To explore full setup protocols, see our comprehensive AI Commerce integration specs. If you need a turn-key framework, consider deploying our high-performance AI SaaS Boilerplate, or work directly with our engineering team on specialized AI Workflow Automation services.