Daan van der Lugt · 4 min
Yes, you can: Shopify has its own connector for Claude and a plugin for ChatGPT that let you ask about the products, orders, customers and sales in your store. For the question of what you actually keep, that is not enough, because your ad spend, marketplaces, carriers and accounting live outside Shopify. For that, your AI needs one model that brings all those systems together, with fixed agreements on what profit is.
Can you connect Claude or ChatGPT to Shopify?
Yes. Shopify has built integrations for Claude, ChatGPT, Gemini, Perplexity and Microsoft Copilot, among others. In Claude it is called the Shopify connector, in ChatGPT the Shopify plugin. You install them from the AI tool, not from the Shopify App Store, and then approve in your Shopify admin which data the tool may access.
According to Shopify, the integration itself costs nothing extra; you pay your Claude or ChatGPT subscription to Anthropic or OpenAI. You connect one store at a time. Shopify's guide: Connecting AI tools to your store.
What can such an integration do?
From the chat you work with your products, collections, inventory, orders, customers, discounts and analytics. You can ask, for example, for your sales over the last 30 days by product, and get tables and charts back from your store's analytics. The integration can read orders, but not change them.
Shopify also lists the limits. A connected AI tool cannot issue refunds, cancel orders or process returns, even with write access. Once you share data, it falls under the AI provider's terms instead of Shopify's. And because AI can make mistakes, Shopify advises you to always check answers and changes.
What does Sidekick do?
Sidekick is the AI assistant inside your Shopify admin. You ask it questions, have it do tasks such as editing products, and have it build reports in Shopify Analytics, for example sales, costs, profit and returns per product. Sidekick presents changes for your review first, and it can also work with third-party apps.
According to Shopify, the connectors for Claude and ChatGPT do not replace your Shopify admin or Sidekick: they are another way in. More: Sidekick in the Shopify Help Center.
What is MCP?
MCP (Model Context Protocol) is an open standard for connecting AI assistants to other systems. Anthropic introduced it and later donated it to the Linux Foundation's Agentic AI Foundation, co-founded by Anthropic, Block and OpenAI. Claude and ChatGPT both support MCP, and the connectors in Claude's directory are built on it.
For you as a merchant it means: any system with an MCP server can be connected to your own AI, and you decide which data it may see.
TermMCPWhich MCP servers does Shopify have?
Besides the connectors, Shopify offers MCP servers for developers. The Shopify AI Toolkit gives coding assistants access to Shopify's documentation and APIs. For AI agents that shop on behalf of a buyer, there are servers for the catalogue, the cart and checkout, built on the Universal Commerce Protocol (UCP). And the Customer Accounts MCP server helps a signed-in customer with their own orders.
Those servers are meant for building and for selling through AI, not for your own management figures. Overview: Shopify for developers.
Can you connect other systems too?
Yes. In Claude you can add your own connectors next to the ones in the directory, through an MCP server on the internet. In ChatGPT you connect apps such as Google Drive under Settings > Plugins, and custom MCP servers are fully available to Business and Enterprise workspaces. What exactly is possible differs per plan and changes regularly: see the guides from Claude and ChatGPT.
Each connection gives your AI the data of one system, in that system's shape and with that system's definitions.
Why doesn't your AI know what you earn?
Profit per order depends on costs that sit in different places. Your Shopify orders and the costs you track in Shopify are in Shopify. But your ad spend is in Meta and Google, your sales and commission on bol and Amazon are in their statements, your shipping costs are with your carrier or fulfilment partner, and returns and refunds arrive weeks later.
If you ask an AI which campaign made a profit last month after returns, it has to combine all those sources, attribute orders to campaigns and decide which costs count. If it does that from scratch in every conversation, the answer can differ per question, and your AI and your dashboards stop matching. That is why your AI works better on one model with fixed definitions, where those choices have already been made.
TermProfit after order costsprofit after order costs = revenue − cost of goods − shipping − payment fees − commission − returns and refundsWhat does Ecommy do?
Ecommy connects your shop, marketplaces, ads, inventory, shipping, payments and accounting, and runs everything through one model with profit at its core: profit after cost of goods, after order costs and after marketing, on your definitions. Your own AI, such as Claude, ChatGPT or Copilot, queries that model through MCP. That way your AI gets the same numbers as your dashboards, and you decide what it may see.
TermProfit after marketingprofit after marketing = profit after order costs − ad spend