Walmart Opens Its Platform to AI Shopping Agents via Model Context Protocol

Walmart announced that starting next month shoppers can attach their own AI agents to the retailer’s website through a new "Agent Profile" system. Each agent receives a dedicated profile inside the user’s account, and by default every purchase requires explicit approval. Users may also set a budget ceiling — for example, $100 — and review agent‑driven orders separately from their own.

The retailer built an "Agent Gateway" based on the Model Context Protocol (MCP), an open standard that lets AI agents plug into web services without navigating a browser. MCP also prevents agents from accessing a shopper’s full purchase history; they only see the data needed to fulfill the current task, such as preferred brands or recurring items.

Walmart’s terms require agents to act strictly within the shopper’s intent, forbid changes to payment methods, and mandate that any non‑compliant agent be blocked. The same capability will extend to Sam’s Club, giving wholesale buyers the same automated purchasing option.

Why it matters for GPU and AI infrastructure

AI shopping agents rely on real‑time inference, natural‑language understanding, and secure API orchestration — workloads that scale best on high‑performance GPU clusters. As retailers adopt open protocols like MCP, demand for low‑latency, multi‑tenant GPU cloud capacity will rise, creating new opportunities for providers that can deliver reliable, compliant compute for agent‑driven commerce.