Using ChatGPT with Shopify MCP for Product Discovery and Store Data
“Shopify MCP ChatGPT” usually refers to one of two setups. One makes your products discoverable to shoppers inside ChatGPT, which Shopify handles through its agentic storefronts. The other connects ChatGPT to your own store data through an MCP server, so you can ask questions about products, inventory and orders in a chat.
The two work very differently. The first is mostly a setting in your Shopify admin, while the second needs a ChatGPT plan with Developer Mode, an MCP server and an access token you control. This guide explains both, so you can tell which one you need and set it up safely.
What Shopify MCP Means for ChatGPT
MCP, the Model Context Protocol, is an open standard introduced by Anthropic for connecting AI assistants to outside tools and data. An MCP server exposes a set of tools, such as “search products” or “get order,” and an AI client like ChatGPT calls those tools when a conversation needs them.
Shopify, ChatGPT and MCP meet in three different ways. The table shows which applies to you.
| Goal | How it works | Who sets it up | What ChatGPT can do |
|---|---|---|---|
| Sell to shoppers in ChatGPT | Shopify shares your catalog with AI channels through agentic storefronts | Merchants, from the Shopify admin | Show your products to shoppers and send them to your store to check out |
| Work with your store data in ChatGPT | You add an MCP server connected to your store’s Admin API as a custom app in ChatGPT | Merchants or developers, with an eligible ChatGPT plan | Answer questions about your products, inventory and orders, and make changes if you allow it |
| Build Shopify apps or shopping agents | Shopify’s developer MCP servers and Universal Commerce Protocol (UCP) servers | Developers | Help write and check Shopify code, or power agents that search catalogs and build carts |
Most merchants searching for this topic want one of the first two rows. Selling in ChatGPT needs no MCP setup from you, while working with your own data does.
How Your Shopify Products Appear in ChatGPT
Shopify makes products discoverable in ChatGPT through agentic storefronts. Your products are shared with AI channels through Shopify Catalog, and ChatGPT can surface them when they match what a user is looking for. Purchases happen on your own online store, not inside ChatGPT.
For eligible stores, this is on by default. The setting Allow Shopify to manage for me shares your products with AI channels and enrolls your store in new agentic channels as they launch, so you may already be visible without having changed anything.
Manage or turn off AI channels
Agentic storefronts have their own page in the Shopify admin, under Sales channels > Agentic. There you can see performance across AI channels, the queries your products appear for and recommendations for improving your product data.
To control sharing yourself, turn off Allow Shopify to manage for me. Separate switches then appear for Shopify Catalog access, automatic enrollment in new channels and direct checkout, along with switches for each channel. If you turn off catalog access, Shopify notes it can take up to 7 days before your product data stops being shared.
To keep a single product out of AI channels, you can mark it as unlisted. Be aware that this also hides the product from sitemaps, search engines and your own store search, so it suits products you don’t want found anywhere outside a direct link.
The MCP servers behind it
Shopify builds this on its own MCP servers, which implement the Universal Commerce Protocol. They cover catalog search, carts, checkout and order lookups for AI agents, and they’re aimed at developers building shopping agents rather than at merchants.
If you’ve read older guides, note that the original Storefront MCP endpoint at /api/mcp no longer handles catalog and cart tools. It now hosts the shop policies and FAQs tool, while catalog search has moved to the UCP endpoint at /api/ucp/mcp.
Why product data quality matters
ChatGPT can only describe your products using the information your store provides. Clear titles, complete descriptions, accurate variants and prices, filled-in metafields and up-to-date policies give AI channels more to match against a shopper’s question.
Fixing that data one product at a time is slow in a large catalog. Bulkflow supports bulk export, import and updates for Shopify, WooCommerce and Wix, so you can pull product data into a spreadsheet, fill gaps and standardize values, then send the corrected records back to your store.
How to Connect ChatGPT to Your Shopify Store Data
Connecting ChatGPT to your store means adding an MCP server that talks to Shopify’s Admin API. Shopify’s own developer documentation describes its Dev MCP server for documentation and code help, not one that reads your live store data, so a store-data connection usually relies on a third-party or self-built server.
1. Choose an MCP server
Your choice of server decides how much maintenance and trust the connection involves. The main options trade convenience for control.
| Option | Advantages | Watch out for |
|---|---|---|
| Community open-source server | Free, inspectable code, often covers products, orders and customers | Not maintained by Shopify, may lag behind API changes, and many run only on your own computer |
| Managed integration platform | Hosted, handles authentication and usually offers access controls | Subscription costs, and your store data passes through another company |
| Server you build yourself | Exposes only the tools you want, on infrastructure you control | Needs development time and ongoing upkeep |
ChatGPT connects only to remote MCP servers reachable over the internet. A server that runs locally on your computer needs to be hosted somewhere first, and OpenAI points to its Secure MCP Tunnel for servers on private networks.
2. Create store access with limited permissions
Most store-data servers need an Admin API access token, which you get by creating a custom app for your store. Give that app only the access scopes the task requires, such as read access to products and inventory, rather than full read and write access to everything.
Order access has an extra limit. By default, an app can read only the last 60 days of orders, and reading older orders requires the read_all_orders scope, which Shopify must approve for public apps.
3. Turn on Developer Mode in ChatGPT
Custom MCP connections run through ChatGPT’s Developer Mode, and availability depends on your plan. Business, Enterprise and Edu workspaces get full MCP support, including write actions, which OpenAI describes as beta. Pro users can connect MCP servers with read and fetch permissions only, and apps and MCP aren’t available in the mobile apps.
On Enterprise and Edu, an admin grants access under Permissions & Roles > Connected Data, and each permitted user then turns it on under Settings > Apps > Advanced settings. On Business, each admin or owner enables Developer Mode for their own account from the workspace’s permissions settings.
4. Add the server as an app
Go to Apps > Create, enter the server’s endpoint URL and choose an authentication method if the server needs one. Click Analyze tools, complete any OAuth sign-in, then click Create to save the app as a draft.
Test the draft in a new chat by selecting it from the tools menu or mentioning it in your prompt. In a workspace, an admin or owner then publishes it from the drafts list. ChatGPT uses a fixed snapshot of the server’s tools after approval, so changes to the server don’t apply until an admin republishes the app.
5. Start with read-only questions
Begin with questions you can check against your admin, such as “Which 10 products have the lowest inventory?” or “How many orders did we receive yesterday?” Comparing a few answers with what Shopify shows confirms the server is reading the right data before you rely on it.
What You Can Do Once ChatGPT Is Connected
A connected store turns routine lookups into conversations. The exact tools depend on the server you chose, but common tasks fall into a few groups.
- Product questions. Find products missing descriptions or images, compare variants and check prices.
- Inventory checks. List low-stock or out-of-stock items and spot products that haven’t sold recently.
- Order summaries. Summarize recent orders, look up an order’s status or group sales by product.
- Content drafts. Draft product descriptions or tags from existing product data for you to review.
- Small edits. Update a single product’s details, if the server and your permissions allow write actions.
Large changes are a different matter. Shopify’s Admin API has rate limits, so a chat that updates thousands of products one call at a time is slow and hard to review. For catalog-wide changes, a spreadsheet or bulk import is usually easier to check before anything changes, with ChatGPT helping to find the problems first.
Keep Store Access Safe
An MCP connection gives an AI assistant real access to your business data, so set it up with the same care as any other app with store permissions:
- Use the narrowest scopes. Start with read-only access and add write scopes only for specific tasks you’ve tested.
- Keep confirmations on. ChatGPT may ask you to confirm write actions, and Enterprise and Edu admins can choose which actions an app may perform, with new actions disabled by default.
- Protect tokens. Never paste an access token into a chat or a shared document, and revoke or rotate it if it might have been exposed.
- Vet the server. OpenAI warns that untrusted MCP servers increase exposure to risks such as prompt injection, so check who maintains a server and what it does before connecting it.
- Treat deletes as permanent. Deleting a product or customer through the API can’t be undone with a click, so keep a recent export before allowing destructive actions.
- Mind customer data. Check your plan’s data terms before giving ChatGPT access to customer names, emails or addresses.
Conclusion
Shopify MCP and ChatGPT come together in two practical ways: agentic storefronts that put your products in front of shoppers, and MCP connections that let you work with your own store data in a chat. Knowing which one you need, and keeping access narrow and your product data clean, lets you get value from both without surprises. We hope this guide helps you make sense of Shopify MCP for ChatGPT. If preparing product data in bulk is part of that work, you can explore Bulkflow.
