BigCommerce and Cloudflare Launch WebMCP for AI Shopping Agents

On October 21, 2026, Commerce.com, the company behind BigCommerce, announced the launch of WebMCP together with Cloudflare.

The new technology allows AI agents running in the shopper's browser to interact with an online store through specialized tools: search for products, get information about specifications and availability, select variants, build a cart, and send the shopper to checkout.

The main change is that the AI agent no longer needs to recognize the buttons, fields, and other interface elements of the online store on its own.

Instead, the site provides it with a structured set of available actions.

The new functionality is enabled by default for supported BigCommerce stores, without changing the site design, installing apps, or doing custom development.

At the same time, the shopper stays on the merchant's site, and the final checkout is completed through the online store's existing checkout.

WebMCP: Direct AI Agent Access to Product Search and the Cart

The AI Agent Gets the Store's Tools Instead of Having to Recognize Its Interface

On October 21, BigCommerce introduced WebMCP as a new way for online stores to interact with browser-based AI agents.

A typical AI agent working with a website has to analyze the HTML, determine the purpose of elements, and imitate user actions.

For example, to add a product to the cart, it needs to find the corresponding product card, select a size, identify the Add to Cart button, and click it.

This process depends on the page structure.

If the store changes its design, the placement of elements, or the variant selection logic, the agent may make mistakes.

WebMCP replaces this kind of interaction with structured function calls.

The site declares the available actions and passes them to the AI agent through the document.modelContext browser interface.

For example, instead of browsing pages one by one, the agent can call the search_products function and get search results directly from the product catalog.

It can then use get_product_details to check the available variants and add_item_to_cart to add the selected product.

This makes it possible to use BigCommerce's existing commerce infrastructure without having to reproduce user actions in the site interface every time.

Product Search by Specifications, Price, and Variants

In the official announcement, BigCommerce describes a scenario in which the shopper instructs an AI agent to find a particular product.

For example:

"Find a lightweight jacket in my size for under $150."

The AI agent can query the store's catalog, find suitable offers, and compare them.

It uses data directly from BigCommerce.

The agent receives information about the product, its variants, its price, and other available specifications.

After choosing a suitable variant, it can add the product to the cart.

For example, the shopper chooses a black jacket in size M.

The AI agent identifies the corresponding product variant and passes its ID to the cart management tool.

As a result, the product appears in the current user's cart.

This differs from a scenario in which an AI assistant simply recommends a product and provides a link to the online store.

WebMCP makes it possible to perform the actions supported by the commerce system directly during the store visit.

Adding Products to the Cart and Sending the Shopper to Checkout

WebMCP supports not only search but also cart management.

The AI agent can add products, change quantities, select other variants, and remove items.

After the cart is built, the create_checkout_url operation is available.

It creates a link to go to checkout.

At this stage, the AI agent's role ends: the user confirms the final purchase.

BigCommerce and Cloudflare deliberately keep checkout on the online store's side.

The AI does not get the right to complete payment on its own without the shopper's participation.

The store continues to use its existing checkout and its own rules for payment, shipping, and order processing.

Therefore, WebMCP cannot be considered an equivalent of Native Checkout inside Google AI Mode or Meta AI.

Unlike those scenarios, in the BigCommerce implementation the shopper stays on the merchant's site.

Which Actions Are Available to AI Agents Through WebMCP

Product Search: Searching Products Directly in the Catalog

BigCommerce uses the tools of the already existing Storefront MCP.

One of the main tools is search_products.

It makes it possible to search for products by keyword and get a structured list of results.

For example, the agent can request products from a certain category that match the shopper's need.

To narrow down the choice, get_product_details is used.

This function returns information about the product, including available variants and images.

For example, for footwear the agent can get information about available sizes, and for clothing, about colors and options.

An additional tool, related_products, is also provided.

It helps find complementary or similar products.

For example, when a camera is selected, the agent can suggest compatible accessories if the corresponding products are linked in the catalog.

This capability is of interest for Cross-Selling.

However, recommendations are limited to the information the store provides. WebMCP does not guarantee that the AI will determine the compatibility of all products on its own.

Cart Management: Adding, Changing, and Removing Products

BigCommerce provides several tools for cart management.

add_item_to_cart adds the selected product. If there is no active cart, one is created automatically.

update_cart_item makes it possible to change the quantity, variant, or additional product options.

remove_item_from_cart removes an item.

create_checkout_url generates a link to complete the purchase.

For example, the shopper asks the AI:

"Add two size L T-shirts and one size M."

The agent can identify the corresponding variants, add them to the cart, and show the result.

Then the shopper clarifies:

"Replace one size L T-shirt with size XL."

To perform this operation, the agent uses the cart update tool.

This approach makes it possible to interact with the store in natural language without using the familiar product selection interface every time.

At the same time, the availability of individual actions depends on the store configuration, product restrictions, and the user's current permissions.

C2PA: Additional Tools for Verifying Image Provenance

In addition to the commerce tools, BigCommerce added two functions for working with images to WebMCP.

scan_images_c2pa makes it possible to find and check provenance information for images placed on the page.

inspect_image_c2pa is used to examine the corresponding metadata.

These tools work with C2PA, a metadata standard designed to record the provenance and edit history of digital content.

For example, an AI agent can check whether a product image contains provenance information.

These functions should be viewed as an additional content verification mechanism, not as a system for confirming the authenticity of the products being sold.

Cloudflare: WebMCP Is Deployed Without Changing the Online Store's Code

The Script Is Added on Cloudflare's Side, Not in the Store Template

On October 21, Cloudflare explained the technical implementation of the joint project.

BigCommerce already uses Cloudflare infrastructure to serve a large number of online stores.

This made it possible to deploy WebMCP directly at the web page delivery level.

Cloudflare adds a small JavaScript module to the store's HTML response.

As a result, the user's browser is able to discover special tools for the AI agent.

The merchant does not need to edit the HTML or install additional JavaScript in the template.

Processing of requests to commerce functions remains tied to BigCommerce's existing MCP server.

Cloudflare acts as an infrastructure intermediary that passes on the description of the tools and the calls to them.

WebMCP Bridge: Connecting Tools Inside the Browser

The technical implementation uses a special WebMCP Bridge.

It is loaded from the online store's domain and registers the available tools in the browser.

For example, the bridge can call the store's MCP Endpoint at the /api/mcp path.

After registration, the AI agent can discover the available functions.

Instead of analyzing the web page elements on its own, it receives a structured description of them.

For example:

• search_products: product search.
• get_product_details: getting specifications.
• add_item_to_cart: adding a product.
• update_cart_item: changing the cart.
• create_checkout_url: going to checkout.

This mechanism reduces the AI agent's dependence on the HTML structure and the specific design of the online store.

If the store changes the placement of buttons or the look of product cards, the commerce tools themselves can continue to work unchanged.

WebMCP Does Not Replace Server-Side MCP

It is important to distinguish between two technologies that BigCommerce uses together.

Server-Side MCP allows an external AI assistant to connect directly to the online store's server.

This scenario can be used, for example, when the user works with a compatible AI client and gives it access to the store's tools.

WebMCP is designed for AI agents running directly in the shopper's browser.

In this case, the agent discovers the tools when visiting the site.

Both technologies use the shared infrastructure of BigCommerce's commerce functions.

Therefore, the store does not need to separately develop one set of tools for external assistants and another for browser agents.

But the automatic enabling of WebMCP does not mean that Server-Side MCP is already activated.

That is a separate setting.

How to Connect WebMCP in BigCommerce

Step 1. Check Whether the Store Supports WebMCP

According to BigCommerce documentation dated October 21, 2026, WebMCP is enabled by default for:

• All stores using Stencil Storefront.
• Catalyst stores hosted on Cloudflare.

For custom Headless Storefronts running on Cloudflare, separate enablement through BigCommerce Support is available.

In other configurations, availability needs to be checked individually.

You do not need to install WebMCP yourself on a standard supported store.

However, there is an important nuance.

By default, WebMCP may provide only image verification tools.

For AI agents to be able to search for products and work with the cart, you need to activate Storefront MCP.

Step 2. Open the MCP Settings

In the BigCommerce control panel, go to:

This section contains the settings for Server-Side MCP and WebMCP.

Find the WebMCP block.

It shows the status of the feature and the list of tools available to browser-based AI agents.

If Server-Side MCP is turned off, the list may contain only the standard image analysis tools.

This does not mean that WebMCP was installed incorrectly.

For the commerce tools to appear, you need to complete the next step.

Step 3. Activate Storefront MCP

On the MCP settings page, find the Storefront MCP toggle.

To activate it, you need to turn it on and accept the terms of use.

According to BigCommerce documentation, Storefront MCP must be enabled by the store owner.

After confirmation, the system saves the record that the terms were accepted.

Next, you need to select the corresponding storefront.

BigCommerce provides a separate MCP Endpoint for each Storefront.

After activation, it may take up to ten minutes for the server to become fully available.

Key point: when Storefront MCP is activated, its tools are automatically synced with WebMCP.

This allows browser-based AI agents to use the same product search and cart management capabilities.

Step 4. Check the Available Tools

Go back to the WebMCP card in the MCP settings.

The Synced from Storefront MCP block should show a list of synced commerce tools.

Depending on the store configuration, it may include product search, getting specifications, changing the cart, and creating a checkout link.

You need to make sure the tools are displayed correctly.

Special attention should be paid to stores with several storefronts, different currencies, and special pricing rules.

The AI must use the data of the specific storefront the shopper is on.

Step 5. Test How WebMCP Works

BigCommerce provides technical instructions for testing.

First, make sure that WebMCP Bridge is present on the store's pages.

To do this, a developer can check the HTML for the /.webmcp/bridge.js script.

To test in Chrome, you need to enable experimental WebMCP support.

In the browser, open:

After enabling the flag, you will need to restart Chrome.

Then open the store's site.

In Chrome DevTools, you can go to Application → WebMCP and view the tools detected on the page.

The WebMCP: Model Context Tool Inspector extension is also available.

It makes it possible to view tool names, their parameters, and execution results.

For example, you can check:

• Whether the agent finds existing products.
• Whether it correctly identifies variants and SKUs.
• Whether it displays product information correctly.
• Whether it can add the selected variant to the cart.
• Whether the cart persists when moving between pages.
• Whether a correct checkout link is generated.

For tests, it is recommended to use a separate cart that is not linked to a real order.

If the tools do not appear, you need to check that Storefront MCP is enabled, that the browser supports WebMCP, and the Content Security Policy settings.

How to Turn Off WebMCP

BigCommerce allows you to opt out of WebMCP yourself.

To do this, open:

In the WebMCP card, turn off the corresponding toggle.

Then confirm the action in the Disable WebMCP window.

According to BigCommerce documentation, turning it off can take up to several days.

If you later need to turn WebMCP back on, you need to contact BigCommerce Support.

Turning off WebMCP does not make the store unavailable to regular shoppers.

It stops providing the corresponding browser tools but does not prevent AI agents from analyzing the site's public HTML in other ways.

Limitations to Keep in Mind

WebMCP Is Currently Supported by a Limited Number of Browsers

On October 21, Cloudflare specifically emphasized that WebMCP is at an early stage of development.

At the time of the announcement, it is an experimental browser standard, not a feature guaranteed to be available to every internet user.

Testing in Chrome requires enabling experimental settings.

Chrome Origin Trials and use in certain Agentic Browsers are also available.

The BigCommerce control panel lists preliminary support for Chrome and Edge.

Therefore, it would be incorrect to claim that after WebMCP is enabled, all store visitors will be able to use AI shopping.

Actual use depends on the shopper's browser, the available AI tools, and support for the corresponding protocol.

The Cart Is Available to AI, but the Shopper Confirms Payment

WebMCP does not give the AI agent the right to pay for products on its own.

The AI can find a product, select a variant, add it to the cart, and generate a checkout link.

However, the last step of the purchase is performed by the user.

This is an important difference from certain Agentic Checkout models in which the AI assistant interacts directly with the payment infrastructure.

In the BigCommerce implementation, the user stays on the merchant's site and retains control over completing the purchase.

AI Acts Within Existing Permissions

WebMCP uses the shopper's current browser session.

This makes it possible to take into account their sign-in, cart, and available purchase terms.

At the same time, the agent must not get permissions that the corresponding user does not have.

For example, a visitor who is not signed in will not be able to access restricted B2B capabilities just because they use an AI agent.

In addition, using the tools does not mean that the AI automatically receives the shopper's payment card details.

Operations are performed within the designated commerce infrastructure.

Integration Errors Should Not Disrupt Normal Store Operation

BigCommerce and Cloudflare have provided a fallback scenario.

If the browser does not support WebMCP or the MCP server is temporarily unavailable, the special script stops running.

The store's regular interface continues to work.

For businesses, this is important: introducing AI tools should not make the main purchase process dependent on an experimental browser standard.

WebMCP for B2B Stores

Using the Individual Prices and Terms of a Signed-In Buyer

BigCommerce is developing MCP not only for regular E-commerce but also for B2B stores.

Server-Side MCP supports working with signed-in users through Storefront Session Sync.

This mechanism makes it possible to take into account the data of the shopper's current session.

For example, a B2B customer may have individual prices and access to a certain assortment.

With correct session sync, MCP tools use the corresponding buyer context.

Thus, the AI assistant can work with the products and terms available to that specific customer.

It is important that WebMCP provides browser access to already permitted tools and is not a standalone system for managing corporate permissions.

Shopping Lists and Quotes for Corporate Buyers

A separate set of B2B Storefront MCP tools supports working with corporate purchase lists and quotes.

With BigCommerce B2B Edition and the corresponding permissions, AI agents can use tools to create and change Shopping Lists.

Operations with Quotes are also provided.

For example, a corporate buyer can ask the assistant to prepare a list of supplies for a repeat order.

The AI agent can add the necessary items, change quantities, and prepare the list for approval.

If the user's permissions allow it, the agent can also use approval tools.

For quotes, there are functions for viewing them, creating requests, and interacting with a sales manager.

However, the availability of each B2B tool is determined by the store configuration and the buyer's permissions.

Users from the same company may have different capabilities.

Therefore, launching WebMCP does not by itself mean that all browser agents automatically get access to restricted B2B processes.

Practical Value for E-commerce

Reducing AI Agents' Dependence on the Site Interface

The main advantage of WebMCP is the ability to work with the online store without imitating numerous clicks.

This is especially important for stores with a large assortment, filters, product variants, and complex navigation.

If the catalog provides structured search and selection tools, the AI can find the right offers faster.

At the same time, effectiveness depends on the quality of the product data itself.

Incomplete specifications, incorrect variants, and outdated product information will limit the effectiveness of AI product selection.

Simplifying the Choice of Products With Many Variants

WebMCP is especially relevant for categories in which the shopper needs to choose several parameters.

For example:

• Clothing and footwear: size, color, availability.
• Electronics: configuration and specifications.
• Sporting goods: intended use and size.
• Home goods: dimensions and finish options.
• B2B equipment: specifications and quantity.

The AI assistant can use product specifications when making recommendations.

For example, the shopper can ask it to find suitable products in a certain price range and add the best option to the cart.

However, WebMCP is not a system for expert product evaluation.

If the choice requires information that is not in the catalog, a correct AI answer may require additional information.

New Cross-Selling Opportunities

The related_products tool creates an additional scenario for recommending complementary products.

For example, after a camera is chosen, the AI assistant can suggest compatible accessories from the catalog.

For the merchant, this is an opportunity to present additional products directly during AI product selection.

However, an increase in average order value is not guaranteed.

The actual result depends on the quality of the links between products, the relevance of recommendations, and shopper behavior.

Keeping Sales Inside the Merchant's Own Store

WebMCP does not require moving checkout to an external platform.

The shopper stays on the merchant's site.

This allows the company to keep its existing mechanisms for payment, shipping, order processing, and customer service.

For an E-commerce business, this is especially important when it uses its own loyalty programs, complex shipping rules, and additional services.

At the same time, support for AI agents does not mean an automatic increase in conversion.

Results need to be checked against order data and shopper behavior.

What Changes for Performance Marketing and Analytics

WebMCP potentially creates a new scenario for how a shopper interacts with an online store.

The user can find a product and add it to the cart through an AI agent rather than through the site's usual elements.

The transaction is still completed in the store.

This means that standard E-commerce analytics retains its fundamental role.

However, you need to check separately how the AI agent's actions are reflected in analytics events.

For example, if the cart is changed through an MCP tool, you cannot automatically assume that all the usual add_to_cart browser events will be sent exactly as they are when a button on the site is clicked.

The event needs to be checked in the store's actual implementation.

For GA4 and advertising systems, it is important to ensure consistency between the cart state, analytics events, and completed orders.

Special attention should be paid to:

• Tracking add-to-cart actions.
• Tracking checkout starts.
• Recording purchases.
• Preserving traffic sources.
• Avoiding duplicate conversions.
• Passing revenue correctly.

At the same time, the official BigCommerce announcement does not contain confirmed data on the impact of WebMCP on Conversion Rate, AOV, or ROAS.

Therefore, it is too early to talk about an increase in commercial performance.

WebMCP Availability and Technology Status

On October 21, 2026, BigCommerce announced that WebMCP is enabled by default for supported stores.

The technical documentation specifies that this applies to Stencil and Catalyst Storefronts hosted on Cloudflare.

Certain Headless implementations require additional enablement.

The commerce tools become available to browser-based AI agents after Storefront MCP is enabled.

Server-Side MCP itself moved from beta to general availability on October 8, 2026.

WebMCP remains an experimental browser technology with limited support.

Therefore, the launch on BigCommerce's side does not mean that all browsers and AI assistants are already able to use the new tools.

Stores should separately check the availability of features, the correctness of the connection, and the AI's behavior when working with the cart.

BigCommerce: What the Update Means for Online Stores

BigCommerce is an E-commerce platform for building and managing online stores that supports B2C and B2B sales, multichannel commerce, and integrations with external systems.

The launch of WebMCP together with Cloudflare expands the capabilities of its commerce infrastructure for interacting with AI agents.

The main change is providing structured access to product search, getting specifications, and working with the cart directly from the shopper's browser.

For supported stores, the basic WebMCP infrastructure is already enabled, and the commerce tools become available after Storefront MCP is activated.

The merchant retains control over the catalog, prices, checkout, and order processing.

The update is of greatest interest to E-commerce companies with a large assortment, complex product variants, and corporate customers.

The practical priority for such stores is to check that MCP is activated, test the available commerce tools, and make sure that AI agents' actions are correctly reflected in the cart and the analytics system.

Sources

1. Commerce.com. Commerce and Cloudflare Bring AI Shopping Assistants to BigCommerce Merchants’ Online Stores with WebMCP.
2. BigCommerce. WebMCP: Bringing Browser-Based AI Agents to BigCommerce Storefronts.
3. Cloudflare. How Commerce.com and Cloudflare Are Getting Storefronts Ready for AI Agents.
4. BigCommerce Developer Changelog. WebMCP for Storefronts.
5. BigCommerce Documentation. WebMCP.
6. BigCommerce Developer Changelog. BigCommerce MCP Server Is Generally Available.
7. BigCommerce Documentation. MCP Server Overview.
8. BigCommerce Documentation. B2C Storefront MCP.
9. BigCommerce Documentation. B2B Storefront MCP.
10. BigCommerce Documentation. Storefront Session Sync.