How NorthShelf's online store answers inventory questions in real time

NorthShelf's dev team spent one day writing a small MCP server that connects to their existing inventory and order management system. They exposed three tools: check_inventory(sku, size, color), get_order_status(order_id), and get_active_promotions(). Then they pasted the server URL and a bearer token into the AgentNDX dashboard under AI Agent → Live Data.

Now when a shopper asks 'Is the Summit jacket still in stock in slate gray, women's medium?', the agent calls check_inventory in real time. It doesn't guess. It doesn't send them to the product page. It answers: '4 units left in that size and color — these have been moving fast today. Want me to hold your spot while you check out?' The whole thing takes two seconds.

When a returning customer asks about a package, the agent calls get_order_status with their order number. It reads back the tracking update, the estimated delivery date, and whether there's a delay. The support inbox stops filling up with order status emails.