How Commerce.com and Cloudflare Enable AI Agents to Shop Smarter on Storefronts

Key takeaways

• Traditional websites are not designed for AI agents, causing slow and error-prone shopping experiences.
• Google’s experimental WebMCP standard provides structured shopping tools for AI agents directly on storefronts.
• Commerce.com and Cloudflare partnered to deploy WebMCP across nearly 100,000 stores with merchant opt-out control.
• AI agents can now perform product searches and cart actions reliably within the shopper’s browser session without compromising merchant control.
• The integration includes safe fallbacks and customizable server paths to ensure seamless operation for all shoppers.

What challenges do AI agents face when shopping on traditional websites?

Websites have traditionally been built with human users in mind, focusing on visual layout, usability, and direct interaction. However, AI agents that automate shopping do not interact with websites as humans do; instead, they rely on interpreting the site's HTML structure to identify elements such as buttons, input fields, and product listings. This process involves guesswork and heuristics, which can be fragile and inefficient.

For example, an AI agent trying to add a product to a cart might scan the page's HTML to find a button labeled "Add to Cart". If the website changes its theme or layout, even slightly, the button's HTML structure or class names might change, causing the agent to fail or misinterpret the page. Additionally, obstacles like cookie consent banners, variant selectors (e.g., size or color dropdowns), and bot detection challenges complicate the automation further.

This guesswork leads to slower performance, higher computational costs, and frequent errors, resulting in a poor shopping experience for users relying on AI agents. The fragility of such systems limits the scalability and reliability of AI-driven e-commerce.

What is WebMCP and how does it improve AI shopping?

WebMCP (Web Machine Commerce Protocol) is an experimental browser standard developed by Google to address the challenges AI agents face when shopping online. Instead of requiring AI agents to parse and guess page elements, WebMCP embeds structured shopping tools directly into storefront pages. These tools expose clear, programmatic interfaces such as search_products, add_to_cart, and get_checkout_link.

By calling these tools directly, AI agents bypass the need to interpret complex HTML structures or simulate human clicks. This direct interaction makes shopping faster, more reliable, and less resource-intensive. For instance, an AI agent can issue a command like search_products({ query: "trail running shoe", max_price: 140, size: 10 }) to immediately get relevant product results without scanning the page.

Currently, WebMCP is in an early preview phase and is accessible behind experimental Chrome flags (chrome://flags/#enable-webmcp-testing), Chrome Origin Trials, some agentic browsers, and developer environments such as Cloudflare Kitesurf. These environments allow developers and early adopters to test and integrate WebMCP features as the standard evolves.

How did Commerce.com and Cloudflare implement WebMCP at scale?

Commerce.com is a flexible commerce platform hosting nearly 100,000 storefront domains on Cloudflare’s infrastructure. When WebMCP became available in preview, Commerce.com wanted to enable it across all their stores simultaneously without requiring merchants to alter their website themes or code.

To achieve this, Cloudflare developed an edge implementation that automatically injects a small script into storefront pages. This script activates WebMCP tools on the client side, enabling AI agents to discover and use them seamlessly. The injection happens at Cloudflare’s edge servers, ensuring minimal latency and no need for merchant intervention.

Merchant Control and Opt-Out Mechanism

Recognizing the importance of merchant autonomy, Commerce.com and Cloudflare created an opt-out system that allows individual merchants to disable WebMCP for their stores via a simple dashboard toggle. This opt-out is enforced at Cloudflare’s edge through Configuration Rules and custom hostname metadata. When a merchant opts out, Cloudflare serves the storefront without injecting the WebMCP script, ensuring zero additional latency or impact on the user experience.

This approach balances broad deployment with respect for merchant preferences, allowing Commerce.com to provide AI shopping capabilities while preserving control for each store owner.

Technical Adaptations

During implementation, Commerce.com and Cloudflare addressed practical details such as custom MCP server paths. While the WebMCP toolpack initially expected the MCP server at /mcp, Commerce.com’s backend MCP server runs at /api/mcp. Cloudflare’s edge injection script was updated to accommodate this by passing the custom MCP URL path, ensuring compatibility across all stores.

Additionally, safe fallbacks were implemented so that if the MCP server is offline or the customer’s browser does not support WebMCP, the script gracefully exits and the site loads normally, preserving the experience for human shoppers.

What benefits do merchants and shoppers gain from this integration?

The integration of WebMCP into Commerce.com storefronts, delivered by Cloudflare, provides tangible benefits:

• For Merchants: They maintain full control over their customer relationships and checkout processes. Since the AI agent operates within the shopper’s active browser session on the merchant’s domain, no credentials or sensitive data are shared with third parties. Merchants can keep their existing checkout flows intact without rebuilding their sites.

• For Shoppers: AI-assisted shopping becomes faster and more reliable. Agents no longer struggle with cookie banners, variant selections, or bot challenges, resulting in smoother product searches and cart management. Because the shopping session stays on the merchant’s site, shoppers retain control over their purchases.

Example Scenario

Imagine a shopper instructing their AI assistant: "Find me a waterproof trail running shoe under $140 in size 10 and add it to my cart." Without WebMCP, the agent might fail to handle pop-ups or misinterpret variant selectors, causing delays or errors. With WebMCP, the agent simply calls:

• search_products({ query: "trail running shoe", max_price: 140, size: 10 })
• add_to_cart({ product_id: "shoe_123" })

The cart updates instantly, and the shopper can proceed to checkout securely.

What technical safeguards ensure reliability and merchant control?

The WebMCP integration relies on several technical safeguards:

• Edge Proxying: Cloudflare acts as an edge delivery layer, proxying tool definitions and actions from Commerce.com’s backend MCP server directly to the browser. This ensures merchants retain full control over their tool logic.

• Configurable MCP Paths: The system supports customizable MCP server paths to accommodate different backend configurations.

• Safe Fallbacks: If the MCP server is unreachable or unsupported by the browser, the injected script exits cleanly, allowing the storefront to operate normally for human shoppers.

• Opt-Out Enforcement: Merchants can disable WebMCP at any time, with Cloudflare enforcing this choice at the edge to prevent script injection and avoid latency impacts.

These safeguards maintain a secure, stable, and merchant-friendly environment for AI-driven shopping.

What should businesses expect going forward?

Cloudflare and Commerce.com are committed to advancing WebMCP as the standard matures. Future developments will focus on enhancing agent tool discovery, improving shopping reliability, and expanding support for diverse commerce scenarios.

Their collaboration aims to foster a safe and efficient ecosystem for AI-driven shopping that benefits both merchants and consumers. As AI becomes increasingly integral to e-commerce, standards like WebMCP will play a crucial role in shaping seamless, agent-friendly storefronts.

Businesses can anticipate ongoing improvements in AI shopping capabilities, broader adoption of WebMCP, and enhanced tooling for managing agent interactions.

What can businesses do now?

Businesses, especially small and growing ones using Commerce.com or similar platforms, should stay informed about WebMCP developments to prepare for AI-driven shopping experiences. Practical steps include:

• Monitoring platform updates and announcements related to WebMCP.
• Reviewing merchant dashboards to understand and manage WebMCP opt-out settings.
• Exploring options for branded domains, short links, and analytics to support AI tool integration.
• Evaluating edge services like those offered by Cloudflare to enhance performance and security.

By proactively engaging with these technologies, businesses can position themselves to leverage AI shopping tools effectively and maintain competitive advantage.

For more details on managing links and campaigns that drive customer engagement, consider exploring FY9’s offerings such as branded domains, analytics, and campaigns.

Sources

• How Commerce.com and Cloudflare are getting storefronts ready for AI agents — Cloudflare, 21 October 2026