Back to Blog
RecipeFeb 8, 202618 min read

Build an AI E-commerce Order Manager with Stripe MCP + Shopify MCP

NT

Nikhil Tiwari

MCP Playground

🍳 MCP Recipe

  • What you'll build: An AI assistant that manages e-commerce operations across Stripe and Shopify using natural language
  • MCP servers: @stripe/mcp (official, remote + local), Shopify Storefront MCP (remote, per-store)
  • Time to complete: 30–45 minutes
  • Difficulty: Intermediate

Running an e-commerce store means constantly switching between Stripe for payments and Shopify for products and orders. This recipe connects both to your AI assistant so you can manage everything in natural language.

You'll be able to ask things like:

  • "Show me today's Stripe balance and recent payments."
  • "Refund the last payment from customer@example.com."
  • "Search the Shopify catalog for 'wireless headphones' and show me the top results."
  • "Create a Stripe payment link for $49.99 for the Premium Plan."

The Two MCP Servers

This recipe uses two separate servers—one for payments (Stripe) and one for storefront/product data (Shopify). Here's what each provides:

Server Package / URL Key Tools
Stripe MCP @stripe/mcp (npm) or remote: https://mcp.stripe.com customers.create, customers.read, paymentIntents.read, refunds.create, invoices.create, invoices.read, invoices.update, invoiceItems.create, subscriptions.read, subscriptions.update, products.create, prices.create, paymentLinks.create, balance.read, disputes.read, disputes.update, coupons.create, coupons.read
Shopify Storefront MCP https://{shop}.myshopify.com/api/mcp (remote, per store) search_shop_catalog (product search, pricing, variants, images)

Note: Shopify also has a Dev MCP (@shopify/dev-mcp) for documentation and GraphQL schema exploration, and a Customer Account MCP for authenticated customer interactions. This recipe focuses on the Storefront MCP for product/catalog access and Stripe MCP for payments.

Prerequisites

Stripe account

With a secret API key (test mode recommended for setup)

Shopify store

Any active store with products

MCP client

Claude Desktop, Cursor, or VS Code

Step 1: Set Up Stripe MCP

Stripe offers two ways to connect: a hosted remote server or a local server via npm.

Option A: Remote Server (simplest)

Stripe hosts an MCP server at https://mcp.stripe.com. Add it to your client config:

Claude Desktop (claude_desktop_config.json):

{
  "mcpServers": {
    "stripe": {
      "url": "https://mcp.stripe.com"
    }
  }
}

Cursor: Settings → Tools & MCP → Add MCP server:

Field Value
Name stripe
Type sse
URL https://mcp.stripe.com

The remote server uses OAuth for authentication—your client will prompt you to authorize with your Stripe account.

Option B: Local Server (more control)

Run with npx and your secret key. You can restrict which tools are available:

npx -y @stripe/mcp --tools=all --api-key=sk_test_YOUR_KEY

Claude Desktop config for local:

{
  "mcpServers": {
    "stripe": {
      "command": "npx",
      "args": ["-y", "@stripe/mcp", "--tools=all", "--api-key=sk_test_YOUR_KEY"]
    }
  }
}

Use test mode first

Use a sk_test_ key (from Stripe Dashboard → Developers → API keys → Test mode) while setting up. Switch to sk_live_ only when you're confident in the workflow. The AI can create real charges and refunds with a live key.

To limit tools (e.g. read-only):

npx -y @stripe/mcp --tools=customers.read,paymentIntents.read,balance.read,disputes.read --api-key=sk_test_YOUR_KEY

Verify Stripe Connection

In your AI client, ask:

  • "What Stripe tools do you have?" — You should see tools like customers.create, balance.read, refunds.create, etc.
  • "What is my current Stripe balance?" — Uses the balance.read tool.

Step 2: Set Up Shopify Storefront MCP

Every Shopify store exposes a Storefront MCP endpoint. No API key is needed—it's a public, read-only endpoint for catalog data.

The URL format is:

https://YOUR-STORE.myshopify.com/api/mcp

Replace YOUR-STORE with your store's subdomain. For example: https://my-coffee-shop.myshopify.com/api/mcp.

Claude Desktop config:

{
  "mcpServers": {
    "shopify-storefront": {
      "url": "https://YOUR-STORE.myshopify.com/api/mcp"
    }
  }
}

Cursor: Settings → Tools & MCP → Add MCP server:

Field Value
Name shopify-storefront
Type sse
URL https://YOUR-STORE.myshopify.com/api/mcp

Verify Shopify Connection

Ask your AI:

  • "Search the Shopify catalog for 'coffee beans'." — Uses search_shop_catalog and returns product names, prices, variant IDs, images, and URLs.

The Storefront MCP is read-only (catalog search). For admin-level operations (creating/editing products, managing inventory), look into community MCP servers like benwmerritt/shopify-mcp which wraps the Shopify Admin API with OAuth.

Step 3: Use Both Together

With both servers configured, your AI can combine Stripe and Shopify in a single conversation. Here are real prompts you can try:

Order & payment lookups

  • "Show my Stripe balance, then list the last 5 payments."
  • "Look up the customer alice@example.com in Stripe and show their recent payment intents."

Refunds

  • "Refund the payment intent pi_3ABC123 in full."
  • "Create a partial refund of $15 for payment pi_XYZ789."

Product search + payment links

  • "Search Shopify for 'wireless headphones' and show me the top 3 results with prices."
  • "Create a Stripe payment link for $49.99 with the name 'Premium Headphones'."

Invoicing

  • "Create a Stripe invoice for customer cus_ABC with a line item of $199 for 'Annual Support Plan'."
  • "List open invoices and tell me which are overdue."

Architecture Overview

When you ask a question, the AI decides which MCP server to call:

1. You Ask

"Refund the last payment from alice@example.com"

2. AI Routes

Calls customers.read → paymentIntents.read → refunds.create

3. MCP Executes

Stripe MCP calls the Stripe API

4. Confirmation

"Refund of $49.99 issued for re_ABC123"

Important Notes

  • Stripe MCP is in public preview (as of Feb 2026). See docs.stripe.com/mcp for current status.
  • Stripe has write tools: refunds.create, customers.create, invoices.create, etc. can modify real data. Use test mode (sk_test_ key) first. Restrict tools with --tools= for read-only usage.
  • Shopify Storefront MCP is read-only: It searches your catalog and returns product info. It does not create or edit products. For admin operations, use the Shopify Admin API or a community MCP server.
  • Remote servers: Both Stripe (https://mcp.stripe.com) and Shopify Storefront (https://YOUR-STORE.myshopify.com/api/mcp) are remote HTTP servers—you can test Shopify Storefront in MCP Playground.

References

Test Shopify Storefront MCP in the Browser

Enter your store's MCP URL and explore catalog tools instantly

Open MCP Playground →

Related Recipes

Frequently Asked Questions

Can Stripe MCP create real charges?
Yes. Tools like refunds.create, customers.create, invoices.create, and paymentLinks.create perform real operations on your Stripe account. Always use a test mode API key (sk_test_) while experimenting. You can also restrict tools with the --tools= flag to allow only read operations.
Can the AI edit products or create orders in Shopify?
Not with the Storefront MCP—it provides read-only catalog search. For admin-level operations (create products, manage orders, update inventory), you need a server that wraps the Shopify Admin API, such as the community shopify-mcp project, which uses OAuth and supports 30+ admin tools.
Can I test these servers in MCP Playground?
The Shopify Storefront MCP (https://YOUR-STORE.myshopify.com/api/mcp) is a remote HTTP endpoint—you can test it in MCP Playground. The Stripe remote server (https://mcp.stripe.com) uses OAuth, so it's better tested in Claude Desktop or Cursor where the auth flow is handled.
What's the difference between Shopify Dev MCP and Storefront MCP?
Dev MCP (@shopify/dev-mcp) is for developers: it searches Shopify docs and explores the Admin GraphQL schema. Storefront MCP is for product/catalog access: it lets AI search your store's products, see prices, and get variant info. This recipe uses Storefront MCP because it accesses real store data.
NT

Written by Nikhil Tiwari

15+ years in product development. AI enthusiast building developer tools that make complex technologies accessible to everyone.