Shopify MCP: connect your store to an AI agent
T. Krause · · updated
"Shopify MCP" can mean two different things. Shopify runs its own MCP servers: the Dev MCP server, which connects a coding agent to Shopify's developer resources, and the commerce servers built on the Universal Commerce Protocol (UCP), which let an agent search catalogs and build carts for a buyer. To give an agent your own store's products, orders and inventory as tools it can call from any MCP client, including a hosted one, you create an app in Shopify's Dev Dashboard, install it on your store, and put its credentials behind an MCP server that turns Admin API calls into tools.
This guide covers Shopify's servers first, so you can tell whether one of them already does the job. Then it connects a store to Prompt TM with the app's client ID and secret, lists the tools the agent gets, and adds the server to Claude Code. The Shopify details below come from shopify.dev as of November 2026.
Shopify's own MCP servers
The Dev MCP server is part of the Shopify AI Toolkit. It is the toolkit's MCP option and connects a coding tool such as Claude Code, Cursor or VS Code to Shopify's developer resources. Shopify describes it as a server that "runs locally and doesn't require authentication", and the Claude Code setup is one command:
claude mcp add --transport stdio shopify-dev-mcp -- npx -y @shopify/dev-mcp@latest
Shopify recommends installing the AI Toolkit as a plugin. The toolkit also lists store management. Per the same page, it can "prepare and run supported store-management tasks through Shopify CLI's authenticated store context, with you choosing when to execute them". That fits a developer working on their own machine with the CLI signed in.
The UCP servers are for agents that act for buyers. Shopify's agentic commerce docs list MCP tools for searching the catalog, building carts, converting them to checkouts and fetching order state. Each request to a store's /api/ucp/mcp endpoint carries an agent profile, which Shopify uses to "verify your agent and apply the right rate limits and tool access".
Storefront MCP is gone. Storefront MCP, at https://{shop}/api/mcp, let agents search a store's catalog and manage carts. Shopify's migration page says its catalog and cart tools "were removed" and that UCP "now provides these capabilities at https://{shop}/api/ucp/mcp".
The Customer Accounts MCP server handles requests from a signed-in customer, "such as checking order status, retrieving order details, or managing account preferences" (docs). The store needs a custom domain, and the app must meet Shopify's protected customer data requirements.
| You want an agent to | Use |
|---|---|
| Help you write Shopify app, extension or Liquid code | Dev MCP server or the AI Toolkit |
| Find products and build carts for a buyer | UCP servers |
| Answer a signed-in customer about their orders | Customer Accounts MCP server |
| Read orders and change products or stock in your store, from a hosted agent or for several stores | An Admin API app behind your own MCP server (below) |
What an agent needs for your store's Admin API
The agent needs an app with the right access scopes installed on the store, an access token for that app, and tools that call the Admin API with the token. Shopify has two kinds of apps for this today.
A Dev Dashboard app uses the client credentials grant. The app "exchanges its own client ID and secret for a token, with no redirect flow to implement". Tokens expire after 24 hours, and you request new ones as you need them. Shopify adds that the grant "only works when the app and the store belong to the same Shopify organization". A merchant can open the Dev Dashboard from the Shopify admin, and Shopify calls apps made there custom apps for the organization's stores.
An admin-created custom app is the older kind, made directly in the Shopify admin. Shopify pre-generates its access token on install. Shopify's page on these apps says "You can no longer create new admin-created custom apps. Existing apps are unaffected and continue to work." The token is shown once, and you can't rotate the app's API key or secret. A new token means uninstalling and reinstalling the app.
Connect a Shopify store to Prompt TM
Prompt TM stores the credentials, gets and renews the token, and serves the Shopify tools at one MCP URL. The agent calls a tool such as SHOPIFY_LIST_ORDERS, Prompt TM adds the token to the request to Shopify, and the agent gets back the result. It never sees the client secret or the token.
1. Create the app in the Dev Dashboard. Follow Shopify's guide: create the app, create a version with its access scopes, and install it on your store. All of Prompt TM's Shopify tools together need read_products, write_products, read_orders, read_customers, read_locations, read_inventory and write_inventory. Leave out the write scopes if the agent should only read. By default, read_orders covers "orders created within the last 60 days", per Shopify's access scope reference. Add read_all_orders for older ones. After installing, open the app's Settings page and copy the Client ID and Client secret.
2. Add an auth config for Shopify. In the Prompt TM dashboard, open Auth configs → New, choose Shopify and the API key method. Or with the API:
curl -X POST https://www.prompt-tm.com/api/v3/auth_configs \
-H "x-api-key: $PROMPT_TM_KEY" -H "content-type: application/json" \
-d '{ "toolkit": { "slug": "shopify" }, "auth_config": { "auth_scheme": "API_KEY" } }'
3. Connect the store. Pass the shop subdomain (the part before .myshopify.com) with the client ID and secret:
curl -X POST https://www.prompt-tm.com/api/v3/connected_accounts \
-H "x-api-key: $PROMPT_TM_KEY" -H "content-type: application/json" \
-d '{ "auth_config": { "id": "ac_..." }, "user_id": "me",
"connection": { "data": { "shop": "my-store",
"client_id": "...", "client_secret": "..." } } }'
Prompt TM requests a token from https://my-store.myshopify.com/admin/oauth/access_token and then reads the shop's details with it to check that it works. If Shopify refuses, the call answers 400 with Shopify's reason, minus the secret, and the account is marked FAILED. One cause Shopify's guide lists is shop_not_permitted, an app and a store in different organizations. A working account is ACTIVE, and the secret and token are stored encrypted. Prompt TM fetches a new token when the old one has less than five minutes left, and once more if Shopify answers a call with 401. If Shopify then refuses the client ID and secret, the account turns EXPIRED.
For an admin-created custom app, send "api_key" with its Admin API access token in place of the two client fields. Prompt TM sends that token as it is and doesn't fetch new ones. To keep the secret out of your terminal, POST /api/v3/connected_accounts/link returns a page on prompt-tm.com with a form for these fields.
4. Create an MCP server. In MCP servers → New, choose Custom, tick Shopify, and under Only some tools pick the tools the agent should see. Copy the server URL.
5. Add it to Claude Code.
claude mcp add --transport http shopify \
"https://www.prompt-tm.com/mcp/mcp_...?user_id=me" \
--header "x-api-key: $PROMPT_TM_KEY"
Scopes, shared configs and common errors for that step are in Claude Code MCP server setup. Other clients are in the MCP client docs. Then ask something the tools can answer, such as "Which orders from this week are still unfulfilled?"
What the agent can do once it's connected
The Shopify toolkit has 11 tools. Eight only read:
| Tool | What it does |
|---|---|
SHOPIFY_GET_SHOP | Returns the store's name, domain, currency, plan, time zone and contact email |
SHOPIFY_LIST_PRODUCTS | Lists products with variants and prices, filtered by status, title, vendor or type |
SHOPIFY_GET_PRODUCT | Returns one product with variants, options, images and inventory item ids |
SHOPIFY_LIST_ORDERS | Lists orders, newest first, filtered by status, payment, fulfillment or date |
SHOPIFY_GET_ORDER | Returns one order with line items, customer, addresses, payment and fulfillment |
SHOPIFY_LIST_CUSTOMERS | Lists customers with contact details, order count and total spent |
SHOPIFY_SEARCH_CUSTOMERS | Finds customers by a query such as email:jane@example.com |
SHOPIFY_LIST_LOCATIONS | Lists inventory locations and their ids |
Three change the store:
| Tool | What it does |
|---|---|
SHOPIFY_CREATE_PRODUCT | Creates a product with optional variants, options and images |
SHOPIFY_UPDATE_PRODUCT | Changes a product's title, description, vendor, type, tags or status |
SHOPIFY_ADJUST_INVENTORY | Raises or lowers stock of one item at one location by a relative amount |
The Shopify toolkit page lists every input. For an Admin API path these tools don't cover, the REST proxy, POST /api/v3/tools/execute/proxy, sends a request to that path on your store with the same credentials.
Limit what the agent can touch
The scopes. Shopify's grant guide says the token "always carries all the access scopes that the store granted to your app installation". So the scopes you approve on the store set the limit, whatever the agent asks for. A scope you add later only takes effect after you release a new app version and approve it on the store.
The tool list. A custom MCP server shows only the tools you ticked. To block the rest everywhere, set restrict_to_following_tools on the auth config. Accounts connected through it can then call only those tools, and the proxy answers 403 for them.
Approvals. Set the MCP server's approvals to Ask a person before any call that changes something. A product update or stock change then returns approval_required: true and does nothing yet. The approval link goes to your webhook and to the server's page in the dashboard, never to the agent, so the agent can't approve its own call.
The log. Each call appears under Logs with the tool, the user and the outcome.
The same steps for GitHub, with a token instead of a client secret, are in how to give an AI agent access to GitHub. To try it with your store, create a free project and connect a development store first.