Run Your Shopify Store in Claude: AI Toolkit Setup

In short

The Shopify AI Toolkit connects Claude straight to your store. Install the plugin from Claude Code's terminal (not the chat box), reload plugins, authenticate your store, and ask questions in plain English. Three jobs to run first: an alt-text audit of every image, a page-level conversion report, and a bundle analysis to lift average order value.

The Shopify AI Toolkit is Shopify's official way of connecting an AI model like Claude straight into your store. Once it is installed, Claude can read your analytics, audit your products, find conversion problems and make changes, all from a chat. Shopify's president Harley Finkelstein announced it, and it is the single biggest change to how I run a store since Shopify itself.

Shopify documents three install methods and all three are a bit technical. This is the simplest one, with the error most people hit and how to get past it.

What you need

  • A paid Claude plan. The toolkit runs inside Claude Code, which is not on the free plan.
  • The Claude desktop app.
  • Your store's .myshopify.com address.

Step 1: Install the plugin (and the error everyone hits)

  1. In the Claude desktop app, go to the Code section (not Chat, not Cowork) and start a new session. Pick or create a folder for it, say "Shopify AI toolkit". Switch the model to Sonnet; you do not need the expensive model for setup.
  2. On Shopify's AI Toolkit page, scroll to the plugin section and copy the first command.
  3. Here is where people get stuck. If you paste that command into the chat box, Claude replies that the plugin is not available. The plugin command is a terminal command, and Claude Code has two faces: the desktop app and a terminal underneath it.
  4. Open the terminal with Control + backtick (the key under Escape). Type claude and press Enter. Now paste the plugin command. It reports that the Shopify AI Toolkit marketplace was added.
  5. Copy the second command from Shopify's page (the install command), paste it in the same terminal, and choose "install for you".
  6. Type /reload-plugins. If the Shopify commands still do not appear, quit Claude and reopen it.

Now type a forward slash and the word "shopify" in any Claude Code chat. You will see a long list of Shopify skills: admin, customers, functions, Liquid for themes, and more. Hover over each for a description. Many are for developers, but several are built for store owners.

Step 2: Connect your store

In the chat, type: "Authenticate my Shopify store for the Shopify AI toolkit." It will ask for your store domain (the part before .myshopify.com) and what access you need. I answered "operate and analyse my Shopify data". Click through the browser authentication, confirm it in the chat, and test it:

What is my conversion rate over the last 30 days on my Shopify store?

The first few requests will ask for approvals. Choose "always allow" for the read-only ones and they stop. If Claude hands you commands to run yourself, tell it: "I don't want to run the command. Do it for me and give me the results." You get sessions, orders, sales, conversion rate and a short commentary on what to look at next.

Use case 1: SEO hygiene on every image

Look at all the images on my store, check their alt tags, make sure they are SEO optimised, and bring me back a report of the ones I need to work on.

On my store it audited 34 products and 194 images in a couple of minutes and returned a table: missing alt text, generic file names, too short, too long, duplicates, alt text that referenced the wrong product. Then it offered to generate corrected alt text for every problem image, export a CSV for review and bulk update. I run this every time we upload new images. Doing it by hand, product by product, used to take an afternoon.

Use case 2: Find your best and worst converting pages

I want to know which pages on my store have the best conversion rate relative to the sessions they get. Find the high-converting and low-converting pages, analyse them and report back.

You get a page-level conversion report for the last 30 days, which pages are underperforming, which are overperforming, and where to optimise first. The action is obvious once you see it: send more traffic to the pages that convert, and fix or stop sending traffic to the ones that do not.

Use case 3: Increase average order value

I want to increase my average order value. Run a report on which products people buy more than one of, and which products I should sell together, based on the data you have.

The report lists hero products, products that are bought in multiples, and pairs that appear in the same orders, which is your bundle list. Because Claude can also edit the store, you can follow up with "build these as bundles" and it will.

Why this is different from Shopify's own reports

Pulling data is useful, but Shopify can do that. The difference is that Claude connects to the rest of your stack. Add Klaviyo, Microsoft Clarity, Canva or Google Drive as connectors and the same chat can read your email results, your heatmaps and your creative, then act: create products in bulk, rewrite listings, edit pages, build themes, draft customer replies, and run all of it on a schedule.

Keep a human in the loop. Let Claude propose, review what it wants to change, then approve. We call it 80% AI, 20% human, and the 20% is what stops a confident AI from doing something confidently wrong on a live store.

Want the theme-editing side of this? Read how to edit your Shopify theme with Claude Code. Want the whole system installed with you? That is what EcomOS and the Elite Team do.