MCP ServerHTTP

Shopify MCP Server

Shopify’s hosted MCP is available at mcp.shopify.com and lets models work with the same store data you can reach through the Admin API — products, orders, customers, and fulfilment, depending on the scopes you grant. Create a custom app in the Shopify admin, install it on your store, and pass the Admin API access token to your MCP client to drive ops and support workflows from chat.

Hosted URL

https://mcp.shopify.com/

Suggested model

Claude Sonnet 4.5

MCP Playground runs 30+ models on the same workflow: switch anytime, or use Compare mode to run several in parallel and balance quality vs. cost.

Auth

Shopify custom app Admin API access token (Settings → Apps → Develop apps) with the minimum read/write scopes the agent should have.

What the Shopify MCP server does

How models use it and what it is built for.

The Shopify MCP is aimed at merchant operators, support leads, and developers who want answers without building a bespoke admin UI. A model can summarize yesterday’s sales, list low-inventory SKUs, pull a customer’s order history for a refund question, and draft a buyer-facing message using structured order facts. You stay inside Shopify’s API permission model, so a read-only app cannot change prices even if the model asks.

Tools the Shopify MCP server exposes

Typical tools an AI model can call. Exact names vary by version.

  • Search and list orders, products, customers, and collections (scoped by token)
  • Read inventory, variants, and fulfilment status for SKUs and locations
  • Summarise performance metrics your scopes expose (e.g. sales in a time window)
  • Draft or stage updates to orders or customers when write scopes are granted

Example prompts to try

Copy any of these into MCP Agent Studio after connecting.

  • What were my top 10 SKUs by revenue last week?

  • Which products have under 5 units in stock in the main location?

  • Pull order #1002 and draft a short reply to the customer’s delivery question.

Models on MCP Playground

This is not a single-model product: you get the same MCP connection with 30+ models (Claude, GPT, Gemini, DeepSeek, open-weight, and more), you can switch mid-conversation, and you can open Compare mode to run the same prompt against multiple models at once. The card above is a suggested starting point for this server — not the only choice.

Default pick for Shopify

Claude Sonnet 4.5

Claude Sonnet 4.5 handles long order threads and product catalog questions. For cheap listing-only reads, try Haiku 4.5; compare in Agent Studio.

Try the Shopify MCP server in your browser

Open MCP Agent Studio with the connection pre-filled. Add your token, pick any of 30+ models, and start chatting — no install required.

Try Shopify in Agent Studio

Shopify MCP server — FAQ

Common questions about connecting, scoping and using it safely.

What is the Shopify MCP server?

It is Shopify’s hosted remote MCP that connects AI clients to your store’s Admin API through documented tools, so a model can query orders, products, and customers the same way a private app would, without you building a custom integration for every question.

What token does Shopify MCP need?

Use a custom app in your Shopify admin, install it on the store, and copy the Admin API access token. Scope it narrowly — start with read-only product and order access until you need writes for drafts or metafield updates.

Can the agent change prices or process refunds without asking?

Only if your app has those Admin API permissions. We recommend a read-only or staging store first, then add write scopes and confirm every sensitive action in the client before your automation runs in production.

Does the Shopify MCP server work in the browser and with Claude?

Yes. Any MCP over HTTP with your token can connect. In MCP Playground, open the Shopify template in Agent Studio so the URL and transport are set, then add your key.

Which model is best for Shopify store operations?

Claude Sonnet 4.5 is a solid default for multi-SKU and customer-context questions. For pure tabular list pulls, a smaller model is often enough — run two models on the same question in Compare mode in Agent Studio.

Other MCP servers

More on MCP Playground