Shopify AI assistant
“Store owners shouldn't need ten tabs to answer one question. Let them ask, and let the answer do the work.”
The product
github.com/moezkayy/mcp-shopify-assistant ↗
In numbers
The idea
Running a small online store means jumping between Shopify screens to answer simple questions. How are sales this week? Which orders are late? What's about to sell out?
This assistant lets the owner just ask. Built for a made-up candle brand, Ember & Oak, it connects Claude to the store through a custom MCP server, so a plain-English question becomes live numbers, drafted emails and changes written back to Shopify.
Seven tools built around business questions, not dozens of raw API calls.
What it does
- Weekly report. Revenue, orders and average order value against last week, plus the top sellers.
- Late orders. Finds every order waiting too long, drafts one apology per customer (merging a customer's orders into one email), and tags the orders in Shopify, all in one request.
- Low stock. Products at or below their reorder point, days of stock left, a suggested quantity, and a reorder email ready for the supplier.
- Order lookups and notes, so the owner can check or annotate any order without opening the admin.
How it's built
A Python MCP server talks to Claude Desktop and to the Shopify Admin GraphQL API, with token refresh handled for you. The analysis runs in tested Python, so answers are fast and the same every time.
Two backends: a generated sample store (12 products, 30 customers, 150 orders) so it runs with no store at all, and a real Shopify development store for the live run.
One command installs it into Claude Desktop, and a reset script removes the sample's tags and notes so it can be run again.
Stack
Note
Built for my portfolio, with a fictional candle store and generated data. It is not client work.