Shopify AI Toolkit vs Dev MCP: Which Shopify AI Tool Do You Need?

Shopify now has several AI tools that sound alike and do different jobs: the Shopify AI Toolkit, the Dev MCP server, the UCP MCP servers for shopping agents, and Sidekick for merchants. Pick the wrong one and you lose a day. This guide separates them by audience, shows the install commands, and gives our view on which to use. Checked against Shopify’s developer documentation in October 2026.

The Four Shopify AI Tools in One Line Each

• AI Toolkit. For developers who use a coding agent such as Claude Code, Codex, Cursor or VS Code. It bundles skills with the Shopify CLI to give the agent current docs and API schemas, validate GraphQL, Liquid and extensions, and run supported store-management tasks through the CLI’s authenticated store context, with you choosing when to execute. It became generally available in the Spring ’26 Edition.
• Dev MCP server. A local MCP server that gives a coding agent Shopify’s developer resources. It runs on your machine and needs no authentication.
• UCP MCP servers. Catalog, Cart, Checkout and Order servers for agents that shop on behalf of buyers. They are covered in Shopify UCP: agent profiles and trust tiers.
• Sidekick. The assistant inside the Shopify admin for merchants. Apps can extend it, as explained in Shopify Sidekick app extensions.

Install the AI Toolkit or the Dev MCP

You need Node 18 or higher. There are three install paths, and you pick one.

# Plugin (recommended): bundles everything and updates automatically
claude plugin install shopify-ai-toolkit@claude-plugins-official

# Skills only: no auto-update, refresh with npx skills update
npx skills add Shopify/shopify-ai-toolkit

# Dev MCP server: local, no authentication
claude mcp add --transport stdio shopify-dev-mcp -- npx -y @shopify/dev-mcp@latest

Restart your tool after adding the MCP server. The earlier per-surface skills, such as shopify-admin and shopify-liquid, are deprecated. A single shopify skill now covers the Admin GraphQL API, Functions, Polaris, Liquid, Hydrogen and the CLI. If you installed the old ones, remove them and install the new skill. The AI Toolkit docs list the exact removal command and the install steps for other editors.

Which Shopify AI Toolkit Option Should You Use?

• You build apps or themes with a coding agent. Use the plugin if your tool supports plugins. Otherwise use the skills or the Dev MCP server.
• You want errors caught before deploy. Use the toolkit, because it validates GraphQL, Liquid and extensions against Shopify’s schemas.
• You are building an agent that shops. Use the UCP CLI and MCP servers. The toolkit helps you write the code but is not the shopping interface.
• You run a store and want answers or quick changes. Use Sidekick in the admin.

Shopify AI Toolkit Use Case: Migrating Extensions

Shopify’s own example is migrating checkout and customer account extensions. Extensions on API versions before 2025-10 need to adopt Polaris web components when you upgrade. The work is repetitive: bump the version in shopify.extension.toml, convert React code to Preact where required, replace legacy components, update imports and test locally. With the toolkit installed, you can prompt your agent directly.

Migrate the extensions/my-checkout-extension extension to API version 2026-04.

Shopify is clear that the toolkit does not remove the need to review and test the result. The same approach helps with Functions work like the custom volume discount, where schema-aware validation catches a wrong input field before you deploy.

Our Analysis: Trade-offs to Plan For

Currency beats cleverness. Shopify ships API versions every quarter, and a model trained months ago can suggest fields that have moved. The main value of these tools is that your agent reads current schemas and docs instead of guessing from memory. That matters more than any single feature.

Plugin, skills or MCP? The plugin updates automatically. Skills need a manual refresh. The MCP server runs locally and needs a restart after config changes. Our default is the plugin whenever your tool supports it, so staying current does not depend on you remembering.

Treat store management as write access. Running tasks through an authenticated store context means an agent can change real data. Point it at a development store first, read each command before it runs, and keep production credentials out of agent sessions you do not supervise.

Token efficiency is a claim to measure. Shopify says combining skills with the CLI is highly token efficient. We have not benchmarked it. Compare it against your current setup on your own tasks before you change a team workflow.

Naming is drifting. Earlier coverage calls the shopper-facing endpoint Storefront MCP, served at /api/mcp. Shopify’s current documentation groups shopper-facing tools as Catalog, Cart, Checkout and Order MCP under UCP. Check shopify.dev/docs/agents before building on an older tutorial.

This article is part of our Shopify Agentic Commerce guide.

FAQ about the Shopify AI Toolkit

Do I need the toolkit and the Dev MCP server together? No. Pick one install path. The plugin bundles what you need, and the Dev MCP server is the option for tools without plugin support.

Which editors are supported? The docs list Claude Code, Codex, Antigravity CLI, Cursor, Hermes, OpenClaw, Pi and Visual Studio Code, with some of them plugin only.

Does the Dev MCP server need authentication? No. It runs locally without it.