On this page
Shopify MCP covers several interfaces with different purposes. Shopify’s Storefront MCP migration guide maps the former catalog and cart tools to UCP, while preserving the policies/FAQ tool and Customer Accounts MCP. Choose a surface by the buyer task, then check its schema and authorization. This guide also explains where a customer-facing post-purchase edit workflow fits.
The Short Version of the Technical Layer
MCP (Model Context Protocol) is a tool interface used by AI clients. Shopify’s current commerce documentation also describes UCP (Universal Commerce Protocol) endpoints for catalog discovery, checkout, and order access. A client must use the endpoint’s current schema, authentication, and permissions; endpoint names and capabilities are not interchangeable.
| Surface | What it is for | Current documentation |
|---|---|---|
| Store-specific catalog | Product discovery for one merchant | Storefront Catalog — https://{shop}.myshopify.com/api/ucp/mcp |
| Global Catalog | Product discovery across Shopify | Agent catalog overview |
| Policies and FAQs | A merchant’s policy and FAQ answers | Policy and FAQs tool — https://{shop}/api/mcp |
| Checkout MCP | Checkout session and buyer handoff | Checkout MCP |
| Order MCP | Read supported agent-originated order information | Order MCP |
| Customer Accounts MCP | Authenticated customer order and account requests | Customer Accounts MCP — discovered https://{shop}/customer/api/mcp, OAuth authorization code with PKCE |
| Dev MCP | Developer documentation and schema tools | Shopify AI Toolkit |
The Storefront Catalog documentation lists search_catalog, lookup_catalog, and get_product and requires the agent profile in meta.ucp-agent.profile. Checkout and Order MCP have different access rules and supported operations. For the discovery tools and their limits, see our Shopify Catalog MCP guide.
Our Call: Build Against the Current Documented Endpoint
Map old integrations tool by tool: product discovery uses Storefront Catalog, cart operations use Cart MCP, and checkout uses Checkout MCP. Keep policy queries on /api/mcp. Test each replacement’s request schema and access rules in development before moving traffic.
The catalog endpoint may support product discovery, but it does not by itself prove that an agent can create a checkout, read a buyer’s order, or change one. Those are separate workflows. The AI Toolkit’s Dev MCP is for developers and is distinct from buyer-facing commerce endpoints.
What an AI Agent Actually Does Against Your Store
The sequence depends on the agent and the Shopify interface it uses. A catalog request can search or retrieve product data. A checkout flow may require authentication or a signed request and may send the buyer to review or complete the purchase. Shopify’s Order MCP documents get_order for supported agent-originated orders; it is not a general cross-channel order history API. Read each endpoint’s scope before connecting it to customer support or fulfillment.
Authenticated customer access is a separate integration
Customer Accounts MCP lets an assistant handle the signed-in customer’s order details, status, and account preferences for that shop. Discover mcp_api from https://{shop}/.well-known/customer-account-api, then authenticate with OAuth 2.0 authorization code and PKCE. Setup requires a custom domain, customer authentication configuration, and the appropriate protected customer data access. After authentication, use tools/list to inspect the actual tools and schemas; do not assume arbitrary edits or cross-merchant account access.
Order MCP instead uses agent credentials to read orders placed through that agent. Shopify’s order webhook guide describes UCP lifecycle updates and Shopify-side registration through a partner manager. Our Shopify agentic order operations guide covers delivery and reconciliation.
The Post-Purchase Workflow Is Separate
A product discovery or checkout endpoint does not provide a general order-editing operation. Shopify’s current Order MCP documentation describes reading the state of supported agent-originated orders; it does not edit them. That distinction matters when a customer needs to correct an address, quantity, or variant after checkout.
A store should provide a customer-facing path for changes that are eligible under its policy and fulfillment state. Revize’s customer order-editing app supports address, quantity, variant, and cancellation changes before fulfillment during the merchant-configured edit window. Cross-product switching, product additions, discount application, and store-credit refunds are Pro features. The portal requires the Revize app block on the order status page and is also available from the Edit Order email link. See our Shopify order-editing guide for the native editor’s workflow and limits.
What Shopify Documents Today
These are the documented capabilities to verify against Shopify’s current references:
- Store-specific catalog discovery through the UCP endpoint, with documented catalog tools and an agent profile.
- A separate Global Catalog endpoint for broader product discovery.
- Checkout MCP with its own authentication or signed-request requirements and possible buyer handoff.
- Order MCP with a
get_ordertool for supported agent-originated orders. - Customer Accounts MCP for authorized customer-specific requests and the policies/FAQ tool for public store questions.
- A local developer toolset through Shopify’s AI Toolkit and Dev MCP.
An endpoint name alone does not establish unrestricted store access, universal customer-account history, or order-edit permissions. Use the documented scope of each interface. Our Shopify AI Toolkit guide explains the developer tooling separately.
What We Would Actually Do This Quarter
- Choose the endpoint from the task. Use the current Storefront Catalog docs for product discovery, Checkout MCP docs for checkout, and Order MCP docs for supported order reads.
- Check profile and access requirements. The Storefront Catalog requests include
meta.ucp-agent.profile; check the exact requirements for each other endpoint separately. - Test in a development environment. Validate the request schema, returned fields, authentication, and buyer handoff before planning a customer-facing launch.
- Audit post-purchase separately. Make sure customers have a route to review or request eligible order changes; do not expect MCP to edit a completed order.
- Recheck the docs before release. Shopify’s agent APIs and client support can change; use current primary documentation rather than copied examples.
The Bottom Line
Define which data the client needs and whose authorization grants access. Catalog discovery, checkout, public policy answers, customer accounts, and agent-originated order reads have distinct contracts. Our Universal Commerce Protocol developer guide explains capability negotiation beneath the UCP surfaces.
For merchants, keep customer service and post-purchase editing as explicit parts of the design. Shopify’s documented Order MCP can read supported agent-originated orders; it does not provide a general order-editing flow. A customer-facing edit portal, merchant policy, and fulfillment cutoff remain separate considerations.
Frequently Asked Questions
What is the current Shopify MCP endpoint?
Use /api/ucp/mcp for Storefront Catalog and the documented cart and checkout tools. Public policy queries remain at /api/mcp; authenticated customer requests use the discovered Customer Accounts MCP endpoint. Pick the endpoint that supports the specific task.
Which Shopify endpoint should I use for product discovery?
Use the documented store-specific catalog endpoint for one merchant’s catalog, or Shopify’s separate Global Catalog endpoint for broader catalog discovery. Check the catalog overview for current tools and request requirements.
What is UCP?
UCP is Shopify’s open commerce protocol for documented agent, merchant, credential-provider, and payment-provider interactions. Use the endpoint-specific references because discovery, checkout, and order operations do not share one universal permission model.
Do I need an agent profile?
The Storefront Catalog documentation requires the profile in meta.ucp-agent.profile. Check the documentation for each other endpoint rather than assuming the same requirement or permission applies everywhere.
Can an AI agent complete checkout?
Shopify documents Checkout MCP separately. Its current flow requires authentication or a signed request and may send the buyer to review or complete checkout. Consult the Checkout MCP reference for the supported path.
Is Shopify MCP available on every plan?
Availability and access depend on the endpoint and the store’s configuration. Check Shopify’s current plan and endpoint documentation for the specific integration; do not infer plan access from an MCP example.
How is Global Catalog different from Storefront Catalog?
Storefront Catalog is scoped to one merchant. Global Catalog is a separate discovery endpoint for product search across Shopify. Their endpoints and tool contracts differ.
What does Order MCP do?
Shopify documents a get_order tool for reading the current state of supported agent-originated orders. It is not a general order-history endpoint for all of a customer’s channels.
How is Customer Accounts MCP different from Order MCP?
Customer Accounts MCP uses the customer’s OAuth authorization for that shop’s account and order requests. Order MCP uses agent-scoped access and can retrieve only orders placed through that agent. Discover each tool’s schema before choosing a customer-service operation.
Does Shopify MCP solve post-purchase order changes?
No. The current Order MCP documentation describes order reads, not editing. Customer-facing order changes need a separate workflow and remain subject to the store’s policy and fulfillment state.
How do I test a Shopify endpoint?
Follow the official reference for that endpoint and test the documented request in a development environment. Confirm its authentication, required profile data, returned fields, and any checkout handoff before release.
What did Shopify report about AI search growth?
Shopify reported in June 2026 that orders coming from AI search were up 13x year over year, with 49% higher conversion and 14% higher average order value. Those figures describe Shopify’s AI-search reporting, not the performance of a particular MCP endpoint. See Shopify’s report.
What is the highest-priority implementation step?
Choose one specific task, follow its current Shopify endpoint documentation, and test the complete request and customer experience. Keep catalog, checkout, order-read, and post-purchase edit workflows distinct.
Related reading
How to sell on ChatGPT with Shopify agentic storefronts · How to let customers edit Shopify orders