Shopify AI Toolkit Lets Claude Code Run Your Store

Shopify's free, MIT-licensed AI Toolkit connects Claude Code, Cursor, and Gemini CLI directly to live store data, giving agents real-time access to API schemas, documentation, and store execution.

Shopify AI Toolkit Lets Claude Code Run Your Store

Shopify quietly shipped the AI Toolkit on April 9 - no press event, no keynote, just a post on X from the official account and a line in the developer changelog. The release lands a capability that's been cobbled together from third-party MCP servers for months: a first-party, always-current bridge between your AI coding client and a live Shopify store.

TL;DR

  • Shopify AI Toolkit is free and MIT-licensed, released April 9, 2026
  • Supports Claude Code, Cursor, Gemini CLI, VS Code, and Codex
  • Three install paths: Plugin (auto-updates), Agent Skills, or Dev MCP Server
  • 16 skills covering GraphQL Admin API, Storefront, Liquid, Hydrogen, Polaris, and more
  • Agents can read inventory, verify code, and execute real store operations
  • Requires Node.js 18+

The Problem It Actually Solves

AI coding agents hallucinate Shopify-specific code constantly. Not because the underlying models are bad, but because Shopify's API surface evolves quickly - the training data is always a few versions behind. A Cursor agent asked to write a GraphQL Admin query will often produce something syntactically plausible but semantically wrong: referencing fields that no longer exist, using deprecated mutations, or missing required scopes.

The toolkit addresses this directly. Per the official documentation, the goal is to ensure "your agent works with Shopify correctly, rather than guessing." It feeds agents live documentation and validates code against current schemas in real time, before anything touches the store.

Two Different MCPs

One source of confusion in coverage of this release: the Shopify AI Toolkit isn't the same as the Shopify Storefront MCP, which shipped earlier. The distinction matters.

The Storefront MCP is for AI agents that shop your store - product discovery, cart operations, and checkout flows from a buyer's perspective. It's what powers integrations like Shopify's Agentic Storefronts, where Shopify products surface inside ChatGPT and Microsoft Copilot.

The AI Toolkit / Dev MCP is for AI agents that build and manage your store - editing themes, running GraphQL queries, verifying code, and executing admin operations. Developer tooling, not buyer tooling.

What the Toolkit Contains

The Plugin Layer

The recommended installation path is the plugin, which auto-updates as Shopify adds capabilities. Setup for Claude Code takes two commands:

# Enable the Shopify marketplace
/plugin marketplace add Shopify/shopify-ai-toolkit

# Install the plugin
/plugin install shopify-plugin@shopify-ai-toolkit

For Cursor, it's a one-click install from the Cursor Marketplace. VS Code requires enabling the Agent plugins preview in settings and installing via the Command Palette. All paths result in the same underlying capability set.

The 16 Agent Skills

The toolkit ships 16 individual skills that can be installed selectively. The full list covers the breadth of Shopify development:

SkillScope
shopify-adminGraphQL Admin API read access
shopify-admin-executionStore operations via Shopify CLI
shopify-storefront-graphqlStorefront API queries
shopify-functionsCustom logic at checkout and runtime
shopify-liquidTheme templating
shopify-hydrogenHeadless storefront framework
shopify-customerCustomer Account API
shopify-custom-dataMetafields and Metaobjects
shopify-payments-appsPayments extensions
shopify-devGeneral developer resources
shopify-partnerPartner-facing APIs
shopify-pos-uiPoint of Sale UI extensions
shopify-polaris-*Design system (4 skills)

The shopify-admin-execution skill is the one that raises the stakes. It enables agents to execute real changes through the Shopify CLI's store execute capability - bulk SEO updates, discount creation, product image swaps, inventory adjustments. This is write access to a live store.

A developer coding at a laptop in a bright workspace The toolkit integrates directly into coding environments like Claude Code and Cursor, feeding agents current Shopify API schemas without leaving the editor. Source: pexels.com

The Dev MCP Server

For teams that prefer MCP-only setups, the Dev MCP Server runs locally with no authentication required. The configuration for Claude Code:

claude mcp add --transport stdio shopify-dev-mcp \
  -- npx -y @shopify/dev-mcp@latest

For Cursor, the equivalent JSON goes into settings:

{
  "mcpServers": {
    "shopify-dev-mcp": {
      "command": "npx",
      "args": ["-y", "@shopify/dev-mcp@latest"]
    }
  }
}

The MCP path is more limited than the full plugin - it doesn't include the complete skill set and doesn't auto-update. Shopify recommends it only for MCP-specific workflows.

Requirements and Compatibility

Compatibility

ToolInstall PathAuto-Update
Claude CodePlugin + Skills + MCPYes (plugin)
CursorPlugin + Skills + MCPYes (plugin)
Gemini CLIPlugin + Skills + MCPYes (plugin)
VS CodePlugin + Skills + MCPYes (plugin)
Codex CLISkills + MCP onlyNo

Requirements: Node.js 18+, active Shopify store for execution skills. MCP server requires no auth.

The toolkit itself is free. The only cost is an existing subscription to whichever AI coding client you use.

Context: Shopify's Agentic Push

This release fits into a broader pattern. Tobi Lütke used an AI agent to cut parse times on Shopify's Liquid templating engine by 53%, and the company has been explicit that agents are a first-class concern for 2026. The AI Toolkit is part of the Winter '26 Edition, which shipped over 150 new AI features in total.

Shopify VP of Engineering Farhan Thawar has been direct about the urgency internally: "If you don't figure out how to harness agents in 2026, you'll be behind." Shopify's internal teams report around 20% productivity gains through AI-embedded workflows.

A merchant completing a transaction at a Shopify POS terminal The toolkit's store execution skills extend AI agent capabilities from code generation into live store management, including inventory, discounts, and product updates. Source: unsplash.com

The "14x increase in orders sourced from some sort of agents" figure Shopify has cited in recent months refers to buyer-side agent activity - AI shopping assistants purchasing through Agentic Storefronts. The AI Toolkit is aimed at the seller and developer side: agents that build and maintain the store rather than browse it.

Where It Falls Short

The shopify-admin-execution skill gives agents write access with limited native guardrails at the skill level. Shopify's documentation notes that security relies on owner configuration rather than built-in permission scoping within the skill itself. That's a reasonable call for a developer tool, but it puts the burden on each team to define safe operating boundaries before pointing an agent at production.

Manual skills drift

Skills installed via the manual path (cloning the GitHub repo and running npx skills add) don't auto-update. As Shopify's APIs evolve, manually installed skills fall behind - the exact problem the toolkit is designed to solve. Anyone using this path needs a process for pulling updates or risks the same stale-schema problem they had before.

Windows friction

Cursor on Windows may need alternative configuration compared to macOS and Linux. The documentation flags this without specifics, which suggests it's known but not fully resolved.

MCP vs. plugin gap

The Dev MCP Server exposes Shopify developer resources but doesn't include the full 16-skill set. Teams that want both - MCP integration and the complete skill surface - need the plugin path, not just the MCP server. The two approaches are not equivalent.


The toolkit sits alongside the growing ecosystem of MCP-powered developer integrations, where live context replaces static training data for domain-specific coding tasks. For Shopify developers already running Claude Code or Cursor's agent workflow, the plugin install takes under five minutes and removes the biggest category of Shopify-specific hallucination errors. For teams without a clear agent strategy yet, it's a low-risk entry point - free, reversible, and scoped to a single platform.

Sources:

Shopify AI Toolkit Lets Claude Code Run Your Store
About the author AI Infrastructure & Open Source Reporter

Sophie is a journalist and former systems engineer who covers AI infrastructure, open-source models, and the developer tooling ecosystem.