Cloudflare and Commerce.com Implement WebMCP to Prepare Storefronts for AI Agents

The Challenge of AI Agent Shopping

Current AI agents struggle with online shopping because websites are designed for human interaction. Agents typically parse HTML, guess button functions, and often fail, leading to slow, token-intensive, and fragile shopping experiences that break with minor site changes.

Introducing WebMCP for Structured Agent Interaction

WebMCP, an experimental browser standard from Google, provides AI agents with structured tools directly on a webpage, such as 'search_products'. This allows agents to call specific functions, making online shopping faster, more cost-effective, and more reliable. WebMCP is currently available through experimental Chrome flags, Chrome Origin Trials, some agentic browsers, and developer environments like Cloudflare Kitesurf.

Cloudflare's Edge Implementation

Cloudflare launched a beta edge implementation of WebMCP in August. When activated, Cloudflare injects a script at the edge to automatically make these AI agent tools available, requiring no code changes from site owners.

Commerce.com's Adoption

Commerce.com, a commerce platform managing nearly 100,000 domains, is integrating Cloudflare's WebMCP implementation. Kristina Pototska, Product Management Director at Commerce.com, stated that merchants seek ways to engage with agentic shopping to ensure their stores appear in agent search results, maintain customer relationships, and control the checkout process without rebuilding their sites. WebMCP addresses these needs by providing agents with direct tools for product search, detail checking, cart building, and checkout link generation.

Merchant Control and Partnership

Commerce.com sought a secure method to enable WebMCP without requiring merchants to manually edit themes. The partnership with Cloudflare allowed for edge-level opt-outs, giving individual merchants control over the feature. WebMCP also operates alongside Commerce.com's server-side MCP.