Cloudflare and Commerce.com Prepare Storefronts for AI Agents
Nearly 100,000 domains on Commerce.com are gaining real shopping tools for AI agents through an edge-level implementation built on Cloudflare.

Bridging the Gap Between AI and Online Retail
Artificial intelligence can write code and pass professional examinations, but purchasing a pair of shoes online remains a persistent hurdle. Traditional online shopping agents typically operate directly from a website's HTML, analyzing pages, guessing button functions, clicking, and repeating the process if errors occur. This current workflow is slow, consumes high token counts, and can be easily disrupted by simple theme modifications.
To address these limitations, platforms are beginning to adopt emerging standards designed to streamline machine-driven interactions. More details on these developments are available through updates on <a href="https://blog.cloudflare.com/tag/ai/">AI</a> and <a href="https://blog.cloudflare.com/tag/agents/">Agents</a>.
Deploying WebMCP at the Edge
WebMCP is a browser standard from Google that introduces structured tools directly onto web pages, such as product search capabilities. Instead of forcing an agent to hunt for interface buttons, the system permits direct calls to necessary tools, resulting in faster and more reliable transactions. Because WebMCP remains an experimental proposed browser standard in early preview, it is presently accessible via specific Chrome flags, Chrome Origin Trials, select agentic browsers, and developer environments like <a href="https://blog.cloudflare.com/kitesurf/">Cloudflare Kitesurf</a>.
In August, an <a href="https://blog.cloudflare.com/webmcp/">edge implementation of WebMCP</a> launched in beta. When activated, Cloudflare automatically injects a small script at the edge, allowing these tools to function without requiring site owners to author custom code.

Scaling Across Nearly 100,000 Storefronts
Flexible commerce platform <a href="http://commerce.com">Commerce.com</a> utilizes Cloudflare infrastructure to support its nearly 100,000 domains. Following the preview release, the platform sought to enable the capability universally across its store network.
Kristina Pototska, Product Management Director at Commerce.com, noted that merchants frequently inquire about maintaining visibility in agent-driven shopping scenarios. When consumers direct an agent to locate items, merchants aim to secure inclusion in the results, retain customer relationships, and preserve checkout control without undergoing site redesigns. WebMCP addresses these requirements by supplying agents with precise tools for product searches, detail verification, cart building, and checkout link generation.
Dual Architecture and Merchant Control
The integration operates in tandem with server-side <a href="https://docs.bigcommerce.com/developer/api-reference/mcp/overview">MCP</a> architectures to accommodate varied shopping methods. While server-side MCP caters to agents connecting directly to a store, WebMCP supports agents functioning inside the browser of the shopper.
Under the hood, WebMCP tools execute inside the shopper's active browser session on the specific storefront domain. This design ensures agents view identical pricing and cart contents while preventing third-party credential sharing. Cloudflare functions as the edge delivery layer for backend tool definitions, preserving merchant control over underlying logic.

Implementing Platform-Wide Opt-Out Rules
Enabling the feature across nearly 100,000 domains presented an initial engineering challenge, as WebMCP was originally developed as a binary on-off switch for entire domain zones. To resolve this for platforms hosting merchant vanity domains on shared zones, the integration leveraged <a href="https://developers.cloudflare.com/rules/configuration-rules/">Configuration Rules</a> and <a href="https://developers.cloudflare.com/cloudflare-for-platforms/cloudflare-for-saas/domain-support/custom-metadata/">custom hostname metadata</a>.
Merchants wishing to disable the functionality can toggle an opt-out preference within their dashboard. Cloudflare evaluates these edge rules prior to serving pages, ensuring opted-out storefronts experience zero added latency, no remote lookups, and no script injections.
Future Outlook and Industry Events
Engineering teams continue to refine the deployment by accommodating custom MCP paths and building seamless fallbacks. If an offline server or incompatible browser is encountered, scripts exit gracefully to ensure normal storefront operations for human visitors.
Discussions regarding storefront readiness for automated shopping agents are featured at industry gatherings, such as the panel sessions scheduled at <a href="https://www.cloudflare.com/connect/">Cloudflare Connect</a>.
Sources
- Cloudflare BlogHow Commerce.com and Cloudflare are getting storefronts ready for AI agents
Continue chronologically





