BigCommerce WebMCP: how AI agents shop your store
Rama Adi
Katelin Teen
Last edited November 9, 2026
BigCommerce switched WebMCP on by default for eligible stores on October 21, 2026, so a shopper's browser agent can now search your catalog, fill a cart and hand back a checkout link, with no code from you. The shopper still pays on your store, which means your support queue sees the result, not the agent. I would check the setting this week, then make sure your team can answer the tickets an agent's choices create. eesel reads BigCommerce orders live to answer them.
If you run a coding agent such as Claude Code, the eesel CLI can upload your approved returns and shipping policies and ask test questions before anything goes live. Connecting the store itself still needs someone with admin access.
I'm Rama. I build integrations at eesel, including the plumbing our agent uses to read BigCommerce orders, so this announcement landed on my desk the day it shipped. Below is what BigCommerce actually documented, what it left out, and what I would do on the support side.
What did BigCommerce announce on October 21?
Commerce.com, the Nasdaq-listed parent of BigCommerce, announced a partnership with Cloudflare that lets AI agents in a customer's browser search, compare and shop on BigCommerce stores. Its Support Community post the same day called it "WebMCP Now Available for Eligible Stores."
The press release makes four claims worth keeping:
- It is on by default for eligible merchants, with no setup and no theme changes.
- Checkout, the customer relationship and the brand stay with the merchant.
- Merchants can switch it off in a new settings menu that also manages all of Commerce's AI connections.
- It builds on the BigCommerce MCP Toolkit: the server-side Storefront MCPs serve agents connecting from outside, WebMCP serves agents inside the shopper's browser.
Sharon Gee, senior vice president of product for AI at Commerce, put the pitch in one line: "WebMCP replaces guessing with a handshake." Cloudflare's side is a developer preview of WebMCP that injects a small bridge script at the edge. That wording matters, because "on by default" and "developer preview" are both true at once.
How does WebMCP work on a BigCommerce storefront?
WebMCP is a browser standard where a page registers tools that an in-browser agent can call through document.modelContext. On BigCommerce, Cloudflare adds one script tag to your storefront's HTML at the edge. It loads /.webmcp/bridge.js from your own domain, and if the browser does not support WebMCP the script simply exits.
The shopping tools are not a second system. They are synced from your server-side Storefront MCP, which BigCommerce opened to every live, transacting store on May 11, 2026. For a B2C store the documented tools are:
| Tool | What the agent can do |
|---|---|
search_products | Search the catalog by keyword, with pagination |
get_product_details | Read variants and images for a product |
related_products | Find complementary or similar items |
add_item_to_cart | Add a product, creating a cart if needed |
update_cart_item | Change quantity, variant or options |
remove_item_from_cart | Remove a line item |
create_checkout_url | Return a link for the shopper to pay |
Two more tools are on by default whatever you do: scan_images_c2pa and inspect_image_c2pa, which read content credentials from images. They report credentials and return signatureVerified: false, so they do not prove anything is authentic.
The part I would underline for a support lead is the end of the flow. The agent can build the cart, but a human opens the checkout link and pays.
Which BigCommerce stores get WebMCP?
BigCommerce documents this as "eligible stores," and the eligibility list is short.
| Storefront type | WebMCP status |
|---|---|
| Stencil storefronts | Enabled by default |
| Catalyst storefronts hosted on Cloudflare | Enabled by default |
| Custom headless storefront on Cloudflare | Enabled on request through BigCommerce Support |
| Catalyst hosted elsewhere | Not documented |
| Stores without Storefront MCP turned on | Only the two image-scanning tools |
For business buyers, the Support Community post says supported tools can extend to shopping lists and quotes, depending on your store's configuration and the buyer's permissions. The WebMCP docs only list B2C tools, so I would treat the B2B detail as a pointer to check, not a feature to promise customers. One more gap: the May Storefront MCP post lists authenticated shopping, meaning order history and logged-in pricing, as "coming next." I found nothing documenting it as shipped.
How do I check WebMCP on my store, and switch it off?
You can check it in two minutes, and I would do it before your next campaign.
- In the control panel go to Settings, then API and MCP, then MCP. The WebMCP card shows whether it is enabled and which tools it exposes.
- From a terminal, run
curl -s https://your-store.example | grep webmcp. If you see a/.webmcp/bridge.jstag, it is live. - In Chrome, enable
chrome://flags/#enable-webmcp-testing, relaunch, and open DevTools, Application tab, WebMCP pane to see the tools as an agent sees them.
Two gotchas from the docs. If your domain sets a Content Security Policy that does not allow the bridge script, the browser blocks it and no tools register, and BigCommerce says that header is yours to fix. And opting out, which is the WebMCP toggle plus a Disable WebMCP confirmation, can take a couple of days to fully clear, while re-enabling goes through BigCommerce Support. The docs also warn that browser extensions can read and run WebMCP tools on any page, so test with a cart that is not tied to a real order.
Is WebMCP ready for real shoppers?
Not yet for most of them, and BigCommerce says so itself. Its own docs describe browser support as limited and still changing.
- The spec is a Draft Community Group Report dated November 2, 2026, and the document states it is not a W3C Standard and not on the W3C Standards Track.
- Chrome's documentation calls it a proposed web standard, lists an origin trial from Chrome 149, and says the API is built mainly for local browser workflows with a human in the loop.
- Chrome also suggests a confirmation dialog for sensitive actions such as a purchase.
The public reaction is thin so far. The one reply under BigCommerce's announcement post worth reading came from a shopper-side skeptic:
"agents can build the cart now. humans still bail at "ok but which size". love that the robots convert, just don't forget the people who clicked an expensive ad to get here."
That is one voice, not a trend, but it points at the real question. If an agent picks the size, who owns the mistake?
What does it change for your support team?
WebMCP changes who picks the item, not who answers when it goes wrong. An agent may choose a variant from a product description, add it to the cart, and the shopper pays after a quick glance. When the wrong size arrives, that ticket lands in your queue like any other.
None of these ticket types is new, and that is the useful part. WebMCP exposes shopping tools only, so the post-purchase side stays with your team. Here is how I would prepare for each:
- Wrong variant. Make sure variants and options are clean, because agents read them as data. Variant inventory is where mismatches start, and ticket volume follows.
- Where is my order. This is already the biggest bucket. WISMO automation and AI order tracking answer it from live order data, and the order-status playbook covers the edge cases.
- Return or swap. Write your policy down in plain words. Our guide on AI for refunds explains where to keep a person in the loop, and macros for exchanges help your agents stay consistent.
I would also tag these orders. If your helpdesk supports it, ticket tagging on "agent-assisted" orders gives you your own numbers, which beats guessing while BigCommerce has published none.
How does eesel fit with BigCommerce WebMCP?
eesel does not shop and does not use WebMCP. It is the helpdesk teammate that answers what happens after the sale. It connects to BigCommerce, reads orders and refunds live, and works inside the helpdesk you already run, such as Gorgias or Zendesk. Refunds go through approval workflows, and anything it is unsure about goes to a person through clean handoffs.
The closest named proof I can point to is not a BigCommerce store, so I will say so plainly. STIL Classics sells planners and binders on Shopify and runs support in Gorgias. In the month from 8 August to 8 September 2026, eesel prepared 320 Gorgias tickets, looked up 230 Shopify orders and left 430 draft replies as internal notes. A person checks each draft, does any refund, and sends it. Nothing goes to a customer without that. It is the same shape of work an agent-driven order creates: look up the order first, then write the answer.
If you are early in this, the BigCommerce AI overview and the guide to adding AI to BigCommerce are the places to start, and best AI for BigCommerce support compares the options. I also keep notes on the AI chatbot route and n8n automations for stores that want to build.
What is still undocumented?
I would rather list the gaps than guess at them:
- Usage numbers. BigCommerce has not published how many agent sessions or orders come through WebMCP.
- Eligibility beyond the table above. "Eligible" is not defined for every plan or hosting setup.
- Order and account tools. The documented tools cover search, cart and checkout link. Order history is not listed.
- Attribution. I found no documented way to tell an agent-assisted order from a normal one in reports.
- Opt-out timing. "A couple of days" is the only figure given.
If any of these matter to you, ask BigCommerce Support in writing and keep the answer. Until then, treat WebMCP as a new traffic source with unknown volume, much like the first wave of AI search engines, and watch your own order tags.
Try eesel for your BigCommerce support queue
If agent-driven orders are about to reach your inbox, eesel is the teammate that opens each ticket with the order already checked. It reads your BigCommerce orders and refunds, drafts the reply in your helpdesk, and leaves refunds and exceptions to your team. Try eesel free with 100 credits, no per-seat fees.
