Shopify AI Toolkit 2026: What Developers and Merchants Need to Know

The Shopify AI Toolkit is a developer toolset; Shopify’s Universal Commerce Protocol (UCP) and commerce endpoints are related but separate parts of the platform. This guide describes the current toolkit, the UCP endpoints in Shopify’s developer documentation, and what those interfaces do for store operators. Revize’s analysis found that 5.2% of orders are edited after checkout (Revize, 2026); that is context for post-purchase operations, not a claim about AI orders.

The toolkit may help developers find current Shopify documentation and API guidance from supported AI coding tools. Merchants should evaluate commerce integrations using their own needs, Shopify’s current endpoint documentation, and each agent provider’s behavior. This guide focuses on those verifiable parts.

What Is the Shopify AI Toolkit?#

Shopify’s AI Toolkit provides a plugin, agent skills, and a local Dev MCP server for supported AI development tools. Shopify currently lists Claude Code, Codex, Antigravity CLI, Cursor, OpenClaw, Pi, VS Code, and Hermes for the plugin; OpenClaw, Pi, and Hermes are plugin-only. Check the official AI Toolkit documentation for current setup steps and supported tools. The Dev MCP runs locally and does not require a Shopify store authentication for its documentation and schema tools.

The local Dev MCP exposes development resources through the Model Context Protocol. Store operations use a separate authenticated Shopify CLI context; installing a documentation server does not authorize access to a merchant’s store.

The useful development capabilities are concrete:

• Find current documentation: identify the right API surface and version before generating an integration.
• Inspect GraphQL schemas: check field names, arguments, and return types before drafting a query or mutation.
• Validate code: check generated GraphQL, Liquid, and extension code against Shopify’s schemas, then test runtime behavior locally.
• Manage a store through CLI: prepare a scoped operation, authenticate the selected store, review the query and variables, and choose when to execute it. Mutations require explicit opt-in.

Validation catches schema and component errors; it does not establish that a change fits the merchant’s business rules. Review the diff, requested permissions, and target store before execution.

Shopify’s Commerce MCP and UCP Surfaces#

Shopify documents separate commerce surfaces for catalog discovery, checkout, and order access. They have distinct endpoint patterns, authentication, and operations; none should be treated as a universal agent interface.

1. Storefront Catalog through UCP#

Shopify’s store-specific catalog endpoint is https://{shop}.myshopify.com/api/ucp/mcp, with search_catalog, lookup_catalog, and get_product and an agent profile in meta.ucp-agent.profile. The Storefront MCP migration guide moves catalog/cart tools to UCP but leaves search_shop_policies_and_faqs on /api/mcp. Use the Storefront Catalog reference for the discovery contract.

Do not assume this endpoint covers checkout or order changes. Shopify documents checkout and orders as separate agent surfaces with their own authentication, permissions, and supported operations.

2. Global Catalog#

Shopify documents a Global Catalog endpoint at https://catalog.shopify.com/api/ucp/mcp for catalog discovery across Shopify. Its tools and access requirements differ from a merchant’s store-specific catalog endpoint. See Shopify’s agent catalog overview before building; do not reuse old Storefront MCP endpoint or tool examples.

3. Order MCP — access to agent-originated orders#

Shopify’s Order MCP currently documents a get_order tool for retrieving the current state of orders that originated with the agent. It is not general access to a customer’s entire order history and it does not edit orders. The tool requires the relevant order access and agent profile permissions. Read Shopify’s Order MCP documentation before relying on it for a workflow.

Shopify also documents separate checkout capabilities through Checkout MCP. Authentication or a signed request is required, and the flow may hand the buyer off to review or complete checkout. Consult the Checkout MCP documentation for the current contract.

4. Customer Accounts MCP — authenticated customer requests#

Customer Accounts MCP remains supported. Discover mcp_api through https://{shop}/.well-known/customer-account-api; its endpoint pattern is https://{shop}/customer/api/mcp. OAuth authorization code with PKCE grants customer-specific access for order details, status, and account preferences in that shop. The app needs customer authentication setup, a custom domain, and protected customer data access. Use tools/list to discover supported operations. This is distinct from Order MCP’s agent-originated order scope; neither name implies arbitrary order edits.

Universal Commerce Protocol (UCP): The Layer Under the Layers#

Shopify describes UCP as an open commerce protocol for interactions among agents, merchants, payment providers, and credential providers. Its documentation specifies discovery, checkout, and order-related flows. Implementations still have distinct endpoints, request formats, authentication, and permissions; a UCP label does not mean every client can call every tool. See Shopify’s agent profiles documentation and the endpoint-specific references before implementing a flow.

The protocol’s main areas include:

1. Catalog discovery — find products and retrieve supported product details through the documented catalog endpoints.
2. Checkout — create or continue checkout through the documented checkout flow, which may require authentication or a buyer handoff.
3. Order access — retrieve supported order information under the documented profile and access rules; this does not imply that an agent can edit or refund an order.

For merchants, confirm whether the relevant Shopify catalog endpoint is available and configured for your store. An endpoint’s existence does not guarantee that a particular AI product will use it or surface every item. Review our Universal Commerce Protocol developer guide and the current endpoint documentation, then test with the client you intend to support.

What Merchants Can Verify Now#

1. AI search traffic is growing, but that does not prove MCP caused it. Shopify reported in June 2026 that orders arriving from AI search were up 13x year over year, with 49% higher conversion and 14% higher average order value. These are Shopify-reported AI-search figures, not a forecast for every merchant or evidence that a specific MCP endpoint generated the orders. See Shopify’s June 2026 update.

2. Check the actual post-purchase workflow. Shopify’s documented Order MCP can read supported agent-originated order information; it does not edit orders. Merchants should verify their customer-facing route for correcting eligible orders rather than assume an agent endpoint covers it.

3. Keep product information accurate. Clear titles, options, availability, and policies help customers and commerce systems interpret a catalog. Measure results in your own store rather than assuming a particular data format guarantees ranking or conversion.

Using the AI Toolkit as a Developer or Agency#

Start with one development task: inspect the schema, generate a query, validate it, and review its result on a development store.

Install the Shopify plugin for your supported client, or follow the documented skill or Dev MCP installation path. Plugins update automatically; manually installed skills require updates. Follow the setup instructions for the client you actually use.

For store management, authenticate with Shopify CLI using only the necessary scopes. The store execute command runs a reviewed Admin GraphQL operation against that stored authentication. For example, this reads a development store’s first ten products:

shopify store auth --store YOUR-DEV-STORE.myshopify.com --scopes read_products
shopify store execute --store YOUR-DEV-STORE.myshopify.com --query '{ products(first: 10) { nodes { id title } } }'

Replace the placeholder with your development store. Reads work without enabling mutations; writes require --allow-mutations and the corresponding scopes. Review the complete mutation and variables before opting in, verify the target store, and test the intended behavior before a production change.

Enabling Agent Shopping on Your Own Store#

If you plan to build against a Shopify catalog endpoint, follow the current Storefront Catalog request format, including meta.ucp-agent.profile. A successful product search does not establish checkout or order access. Our Shopify Catalog MCP guide explains the discovery tools and limits.

For a branded agent, use Shopify’s current sample projects and endpoint-specific checkout documentation. Plan a staged rollout around your own catalog, authentication, payment, and support requirements; a sample project is not a production integration checklist. Test customer-visible product details and checkout handoffs before launch.

Where Post-Purchase Infrastructure Fits in an Agentic Stack#

Product discovery, checkout, order tracking, and customer edits need their own supported interfaces. Our Shopify MCP and UCP endpoint guide maps those responsibilities.

Shopify documents Order MCP as a read surface for supported agent-originated orders, with UCP order webhooks for lifecycle updates. It does not provide a general order-editing flow. Revize offers a customer-facing portal for eligible address, variant, quantity, and cancellation changes before fulfillment, subject to the merchant-configured edit window. Product additions, cross-product switching, discount application, and store-credit refunds are Pro features. Staff-side order editing is in development. The portal requires the Revize app block on the order status page and can also be opened from the Edit Order email link. See Revize on the Shopify App Store for product details.

Related reading#

Shopify order-editing statistics · How to sell on ChatGPT with Shopify agentic storefronts · How to let customers edit Shopify orders