BigCommerce Storefront MCP: How to Switch It On

BigCommerce has its own MCP server. Since 11 May 2026, any live, transacting BigCommerce store can switch on Storefront MCP in the control panel. It gives AI shopping agents, such as Claude or Cursor, seven tools to search your catalogue, build a cart and hand the shopper a checkout link. Stores on B2B Edition get 17 more tools for quotes and shopping lists.

This guide shows a store owner how to switch it on, connect an AI client, test it, and where you still need your own work.

What Storefront MCP is

MCP (Model Context Protocol) is the standard way an AI agent calls tools in other software. Storefront MCP is BigCommerce's own MCP server for your storefront. Each storefront gets its own server URL, and any MCP client that can reach a remote server can connect to it.

The seven storefront tools work for guests and for logged-in customers:

search_products: search the catalogue by keyword (at least three characters).
get_product_details: fetch a product's variants and images.
related_products: find products related to one product.
add_item_to_cart: add a product to the cart, creating a cart if there is none yet.
update_cart_item: change the quantity, variant or options of a line in the cart.
remove_item_from_cart: remove a line from the cart.
create_checkout_url: generate a checkout link for the cart.

On B2B Edition stores the same endpoint adds 12 shopping list tools (create, update, submit for approval, approve, reject, duplicate and so on) and 5 quote tools (request a quote, read it, message the seller, and turn it into a checkout link at the quoted prices). Each buyer only sees the tools their Buyer Portal permissions allow.

Shopify takes a different route, with Storefront and Catalog MCP switched on for eligible stores; our Shopify Storefront and Catalog MCP guide covers it.

How to switch it on

You need to be the store owner to do this. Other staff accounts cannot.

Log in to the BigCommerce control panel as the store owner.
Go to Settings. Under API and MCP, select MCP. (BigCommerce's launch announcement pointed to Settings > Early Access; if your control panel still shows that, MCP is there.)
Turn on the Storefront MCP toggle, then review and accept the terms of service. The page then records who accepted the terms and when.
Under Select storefront, choose your storefront and copy its MCP API Endpoint. Each storefront has its own URL, so on a multi-storefront store repeat this for every storefront you want agents to reach.
Wait. It can take up to 10 minutes for the server to become available after you switch it on.

That is the whole store-side setup. Everything after this happens in the AI client, not in BigCommerce.

Is it safe to switch on? The tools let an agent do what a shopper can already do on your storefront: browse the catalogue, build a cart and get a checkout link, and on B2B Edition manage the buyer's own shopping lists and quotes within their permissions. None of them reads or changes your control panel, order management or store settings, and none takes payment.

How to connect an AI client

Share the endpoint with whoever will test it: you, your developer or your agency. The client needs Node.js 18 or later on the computer for the Claude Desktop and Warp setups.

Claude Desktop. Open Settings > Developer > Edit Config and add this to claude_desktop_config.json, with your endpoint in place of the placeholder, then fully restart Claude Desktop:

{ "mcpServers": { "bigcommerce": { "command": "npx", "args": ["mcp-remote", "<YOUR_MCP_SERVER_URL>"] } } }

Claude Code. Run claude mcp add --transport http bigcommerce <YOUR_MCP_SERVER_URL>. Add --scope user to make it available in every project.

Cursor. Add this to .cursor/mcp.json:

{ "mcpServers": { "bigcommerce": { "type": "streamable-http", "url": "<YOUR_MCP_SERVER_URL>" } } }

Warp. Open Settings > AI > MCP Servers and use the same configuration as Claude Desktop.

How to test it before you tell anyone

Treat the first session as UAT. You are checking that what an agent reads matches what a shopper sees on your live store.

Ask the agent to find three products you know well: a bestseller, a product with variants, and one that is out of stock.
Compare each answer with the product page: price, variants, options, stock and images. Note anything that differs.
Ask it to add a product with options to a cart, change the quantity, then remove a line.
Ask for a checkout link. Open it and check the cart matches what you asked for. Stop before payment, or place a test order and cancel it, the same way you would test any checkout change.
Repeat on each storefront if you run more than one.

For the wider picture, run the twelve checks in our agent-ready store audit. They cover what an assistant can read, reach and buy on your store, whatever the platform.

Most problems you find here are catalogue problems, not MCP problems: a missing attribute, a variant with the wrong price, a product that should be hidden. An agent will recommend what it can read, confidently, so fix the data first.

What it does not do

It does not take payment. The agent hands the shopper a checkout link, and the shopper completes the purchase on your own checkout. A logged-in shopper's browser already holds their session, so checkout opens with their account.
It does not show order history. No order status or order history tool is listed for shoppers.
It is a storefront server, not a back-office one. It cannot read or change your orders admin, inventory settings, analytics or store configuration.
Logged-in shopping needs a developer. To let an agent act inside a shopper's logged-in session (so prices and catalogue reflect their account), BigCommerce provides Storefront Session Sync. It is documented for Stencil storefronts and is built into a chat widget on your storefront.
It does not put tools in the shopper's browser. Storefront MCP is a remote server that AI clients connect to. BigCommerce's documentation does not mention WebMCP, the newer standard where a web page registers tools for an agent running in the shopper's browser. That is why a WebMCP inspector in Chrome shows nothing on a BigCommerce storefront.

When you still need your own MCP server

An earlier version of this guide recommended building a gateway MCP server over BigCommerce's Catalog, Carts, Checkout and Orders APIs. For storefront shopping, you no longer need one: BigCommerce's own server does that job, and BigCommerce maintains it.

Your own server still makes sense for the work Storefront MCP does not cover:

Back-office access for your team, such as order lookups, stock checks or reports. That is admin data, so it belongs behind authentication, never on a public shopping endpoint. Our guide to remote MCP and OAuth covers that pattern.
Tools specific to your business, such as a fitting guide, a trade price calculator or a compatibility check, that no platform server will ever ship.

If you build one, the old rules still hold. Start read-only, use a store-level API account with the narrowest scopes, rate-limit tool calls, and never let a tool do something the person using it could not do themselves.

A note on WebMCP

MCP servers serve AI clients such as Claude and Cursor. WebMCP serves agents working inside the shopper's browser on your actual pages. They do different jobs, and Storefront MCP covers only the first. You can add a WebMCP layer to a BigCommerce theme yourself; our WebMCP guide shows how. Browser agents are moving fast: on Shopify they can already work inside the checkout page. Our agent-ready store audit has the dates and what each platform supports today.

Vortex IQ practices this

WebMCP is live on our own site, with 19 tools an in-browser agent can call, 16 of which any MCP client can also reach at vortexiq.ai/mcp. The reference implementation is public on GitHub, and our authenticated remote MCP connector is live in Claude's connector directory.

How Vortex IQ helps

Switching on Storefront MCP is a few clicks. Making sure an agent reads the right price, the right variant and the right stock level is the longer job.

Vortex IQ, the AI workforce for ecommerce, checks the product data an agent will read across your storefronts, flags what is wrong or missing, and prepares the fixes. It proposes. Your team approves. Nothing goes live without your say-so. We work across BigCommerce, Shopify, Adobe Commerce and Magento. Start with a free store audit.

Frequently asked questions

Does BigCommerce have a native MCP server?

Yes. Storefront MCP is BigCommerce's own MCP server, available to every live, transacting store since 11 May 2026. A store owner switches it on under Settings > API and MCP > MCP.

Who can switch on Storefront MCP?

Only the store owner. Other staff accounts cannot enable it.

Can an AI agent complete a purchase on my store?

Not through Storefront MCP. The agent builds the cart and generates a checkout link; the shopper completes payment on your own checkout.

Does it work for B2B?

Yes, on B2B Edition. The same endpoint adds 17 tools for shopping lists and quotes, and each buyer only sees the tools their Buyer Portal permissions allow.

Why can't I see BigCommerce tools in a WebMCP inspector?

Because Storefront MCP is a remote server, not tools registered in the page. An AI client connects to its URL directly. A WebMCP inspector only shows tools a web page registers in the browser.

Do I still need a gateway MCP server?

Not for storefront shopping. You may still want your own server for back-office access or for tools specific to your business, kept behind authentication.

Can Vortex IQ help?

Yes. Vortex IQ checks the catalogue data an agent will read across your storefronts and prepares the fixes for your team to approve. Start with a free store audit.

Key takeaways

BigCommerce has a native Storefront MCP server, available to every live store since 11 May 2026.
The store owner switches it on under Settings > API and MCP > MCP, picks a storefront and copies its endpoint; allow up to 10 minutes.
Seven tools cover search, product details, cart and a checkout link. B2B Edition adds 17 tools for quotes and shopping lists.
Agents do not take payment: the shopper finishes on your checkout.
Test it like any release: compare what the agent reads with your product pages, then fix the catalogue data.
Build your own server only for back-office or business-specific tools, behind authentication.

Sources

Updated October 2026. An earlier version of this guide said BigCommerce did not ship a native MCP server and recommended building a gateway. That was wrong by the time it was published: BigCommerce made Storefront MCP available to every store on 11 May 2026. We have rewritten the guide around it.

BigCommerce Storefront MCP: How to Switch It On and Test It Safely