Step-by-Step Guide to Connect Shopify with Claude in 2026

Most attempts to use Claude for Shopify work the same way: you paste in some product data or theme code, ask Claude to improve it, then manually apply whatever comes back. It gets the job done for one-off tasks, but it doesn't scale, and Claude has no real visibility into your store.

That changes when you connect them properly. Using a Model Context Protocol (MCP) server, Claude gets structured, live access to your store's GraphQL schemas, Admin API, and developer documentation. With that connection in place, Claude can audit themes, run bulk product updates, and introspect your API schema before executing queries, operating on your actual store data rather than a snapshot you pasted into a chat window.

For teams that need more than a native MCP connection offers, CData Connect AI extends this as a managed MCP platform, adding centralized authentication, access controls, and audit logging across every AI-to-store interaction.

Step 1: Prepare your environment for integration

Before Claude can touch your store, the toolchain needs to be right. Skipping this step means debugging authentication failures or silent command errors later.

You'll need three things installed and current:

Claude Code is Anthropic's terminal-focused tool that reads files, executes CLI commands, interacts with APIs, and edits code directly on your machine. It's not a chat interface with a text box; it operates in your local environment, which is what makes store-level operations possible.

Install Node.js first, then add the Shopify CLI:

npm install -g @shopify/cli @shopify/theme

Note: @shopify/theme is no longer a separate package; it is bundled into @shopify/cli in current versions.

If you prefer working in an IDE, the Shopify AI Toolkit also works with Cursor and VS Code extensions that support MCP-enabled agents. Claude Code is the reference path, but it's not the only one.

Step 2: Install and enable the Shopify AI toolkit in Claude

The Shopify AI Toolkit exposes GraphQL schemas, developer documentation, and live store operations to AI assistants. Once active in Claude Code, you get access to purpose-built tools that replace a lot of manual API work.

The recommended installation method is the plugin, which bundles everything into a single install and updates automatically as Shopify releases new capabilities.

Plugin installation sequence (recommended):

1. Enable the Shopify marketplace in Claude Code:

/plugin marketplace add Shopify/shopify-ai-toolkit

2. Install the plugin:

/plugin install shopify-plugin@shopify-ai-toolkit

Alternatively, to connect via the Dev MCP server directly:

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

3. Reload Claude Code so the new tools register correctly.

After reloading, the Dev MCP server provides these tools for Shopify development (among others in the full toolkit):

The full toolkit ships seven tools in total. The three above come with the Dev MCP server path. The plugin installation path gives you all seven, including store execution capabilities. If you installed it via the plugin, you already have everything — the table above is only a partial view of what's available.

With these tools active, Claude queries your live schema, pulls current documentation, and validates changes before they run. Without them, you're working with static context and hoping it matches your store's current API state.

MCP is an open standard for connecting AI agents to business systems through secure, contextual endpoints and pre-built connectors. The Shopify AI Toolkit is one implementation of that standard.

Step 3: Authenticate and connect your Shopify store

The first time you run a store-connected action in Claude Code (connect store or list themes both work), an OAuth browser window opens. Complete the flow, and Claude is authorized to access your store. That authorization carries across sessions, so subsequent work picks up without re-authenticating.

Shopify CLI also supports Theme Access passwords and custom app access tokens as alternative authentication methods if OAuth doesn't fit your workflow.

For advanced operations through community MCP servers (such as GeLi2001's shopify-mcp), you'll need a Shopify access token with specific API scopes. These tokens follow the shpat_... format and are generated from your Shopify Partner or store admin dashboard.

Security practices worth keeping:

• Complete OAuth only in trusted browsers on machines you control.

• Never share access tokens in Slack, email, or version control.

• Audit and revoke app access permissions on a regular schedule. Treat this like rotating passwords.

Step 4: Work safely with development stores and theme backups

Claude Code executes bulk changes quickly, which is useful in production and risky in experimentation. Always work against a development store or backup theme first.

Before any modification session, back up the live theme:

shopify theme pull --theme {live-id} --path ./.backup-{timestamp}

Commit the backup to a version control branch. If something goes sideways, you have a clean rollback point and a diff showing exactly what changed.

When you're ready to share progress, use shopify theme share to generate a preview URL. According to Shopify's documentation, a visitor preview link (URL) is valid for 2 days, or a merchant preview link is valid for 30 days for reviewers with admin access. This lets a client or stakeholder review the actual storefront without touching production. Only push to live after that review is complete.

Work in a development store, validate with a preview URL, push to production only when everything checks out. Keep version control running throughout.

Step 5: Use MCP tools to introspect schema and validate changes

One of the most useful things the Shopify AI Toolkit enables is programmatic schema introspection. In a GraphQL context, introspection means querying the API itself to understand its current structure before writing queries against it. Admin API fields get deprecated; a query built against an older schema can fail silently or return unexpected data.

The workflow:

• Call introspect_admin_schema before crafting any Admin API GraphQL queries. This prevents using deprecated or renamed fields.

• Use Claude's search_dev_docs tool to validate code blocks for bulk products or theme edits before running them.

• Run Shopify's built-in linter:

  shopify theme check

Claude Code can interpret the linting output and fix flagged issues. If the linter flags images missing a loading="lazy" attribute, Claude applies that fix across affected files automatically.

Schema introspection plus linting catches most issues that would otherwise surface as runtime errors in a live store.

Step 6: Preview and push changes to your live store

At this point, your changes are tested in a development theme and validated by the linter. The last step before going live is stakeholder review via preview URL.

shopify theme share

This creates a visitor preview link. Share it with whoever needs to sign off: a client, a merchandising lead, a QA reviewer. Only after that review clears, push to production:

shopify theme push --theme {live-id}

Watch the logs during the push. Claude Code can help interpret error output in real time if something unexpected appears.

Pre-live checklist:

• Theme backup committed to version control

• shopify theme check linting passes

• Preview URL reviewed and approved by stakeholders

• Bulk changes verified on a per-item basis

• Rollback path confirmed

Claude Code can execute bulk operations (changing product descriptions across 40+ products, for instance) in seconds. Verify each major operation immediately after it runs.

Best practices and operational tips for secure and reliable integration

• Keep changes incremental: Avoid vague prompts like "update all my product pages." Specific prompts like "change the product page background color to #F5F5F5" give Claude a bounded task with verifiable output.

• Use version control for everything: AI-assisted tools move fast. Git branches give you the diff history and rollback capability to keep that speed from becoming a liability.

• Know what community MCP servers add: Servers like GeLi2001's shopify-mcp extend the Toolkit with additional agent tools covering collections, draft orders, and bulk operations. Check the repository directly for the current tool count, as it changes with updates. These require correctly scoped API tokens; read the scope requirements before generating access credentials.

• Stay current with toolkit updates: The Shopify AI Toolkit is actively developed. Installing via plugin handles updates automatically. If you installed via agent skills or MCP directly, pull updates manually before starting new projects to avoid schema drift.

• Document your integration workflows: Record which MCP tools you're using for which tasks, the API scopes they require, and your rollback procedure. This pays off when onboarding a new team member or debugging an incident months later.

When you need more than the AI toolkit: CData Connect AI

The Shopify AI Toolkit handles the Claude-to-store connection well. What it doesn't address is a broader IT challenge: how do you govern AI access to live business data at scale, across multiple systems, without building custom integrations for each one?

CData Connect AI is a managed MCP platform that connects AI assistants like Claude to hundreds of enterprise data sources, Shopify included, through a governed, no-code layer. Rather than managing raw API tokens and scopes per connector, Connect AI centralizes authentication, enforces access controls, and provides audit logging across every AI data interaction.

For teams running Claude across Shopify alongside ERP, CRM, and marketing platforms, it provides the enterprise governance layer the native toolkit doesn't include: real-time data access without pipelines, and the audit trail compliance teams require.

Learn how to connect Shopify with Claude using Connect AI in our KB documentation.

Frequently asked questions

Start querying live Shopify data with Claude

CData Connect AI handles connectivity, security, and governance so you don't have to. Start a 14-day free trial and run your first live query in Claude today.