Walmart Opens Its Marketplace to AI Shopping Agents
TL;DR: Walmart unveiled Agent Gateway on October 21, 2026, letting third-party AI shopping assistants browse and buy through a structured data protocol (MCP) instead of scraping rendered web pages. It ships alongside Agent Profiles, which set per-agent spending caps and purchase-approval rules, and Purchasing Rules, which keep payment-card data out of an agent's hands. The move is a direct contrast with Amazon, which has been actively blocking AI agents from its storefront. For brands, the practical takeaway is that agents read structured product attributes, not pretty page layouts — and that rewards sellers whose catalog data and APIs are actually clean.
What Walmart actually shipped on October 21
Agent Gateway is built on the Model Context Protocol (MCP), the open standard that lets AI tools query structured data sources instead of parsing HTML meant for human eyes. Through it, a connected AI agent can search Walmart's catalog, build a cart, and check out on a shopper's behalf, reading product attributes directly rather than guessing at them from a rendered page. Walmart paired the launch with two guardrail layers: Agent Profiles, which let a shopper define what a given agent is allowed to spend and whether it needs approval before completing a purchase, and Purchasing Rules, which keep the agent out of raw payment-card data entirely.
The contrast with Amazon is the part getting the most attention. Amazon has spent much of 2026 actively working to block AI shopping agents from browsing and buying on its site, treating unauthorized agent traffic as something closer to a scraping problem than a feature to build for. Walmart's bet is the opposite: that agentic commerce is coming regardless, and the marketplace that gives agents a clean, sanctioned way in will capture that traffic instead of fighting it.
What this means for brands
If AI agents are going to read your listings the way Agent Gateway intends, they are reading structured fields — title, attributes, variant data, availability — not the visual layout a human shopper sees. A listing that looks fine on a browser but has incomplete or inconsistent attribute data is effectively invisible to an agent making a purchase decision on a shopper's behalf. That is a different kind of catalog hygiene problem than the one most Seller Center teams are set up to catch manually, and it only gets harder across a multi-marketplace catalog where attribute schemas do not line up one-to-one.
For brands running catalogs at any real size, getting ahead of this is less a Seller Center task and more an integration one: pulling product data into a system that can validate and normalize attributes against what an MCP-based agent is actually going to query, rather than hoping the existing listing happens to be complete. That is the kind of work brands are increasingly handing to YuSMP Group's AI integration team instead of trying to retrofit by hand, listing by listing.
There is also a spending-control question every brand should look at now rather than later. Agent Profiles put the cap-setting decision in the shopper's hands, which is the right default, but it also means a brand has no direct lever over how an agent is configured to buy its products — only over whether its catalog is structured well enough to be read correctly in the first place.
What to do now
- Audit attribute completeness, not just listing appearance. Run a sample of your Walmart SKUs through a structured-data check rather than a visual one — missing or inconsistent attributes are what an agent actually trips over.
- Treat this as a data-pipeline question, not a one-time cleanup. New SKUs and variant updates need the same attribute discipline going forward, or the gap reopens within a quarter.
- Watch for Amazon's response. Given the public contrast Walmart is drawing, expect Amazon to clarify its own agent policy; brands selling on both marketplaces may end up managing two different catalog-readiness standards.
