Walmart will let shoppers create a separate profile for each AI shopping agent linked to their account, complete with spending limits and purchase approvals. The feature, dubbed Agent Profiles, rolls out next month.

The move stands in sharp contrast to Amazon, which blocked Meta's Muse agent from its store less than two weeks after its launch, citing customer security and user experience. The two decisions mark an emerging split in how major retailers are approaching agentic commerce: build controlled entry points for outside agents, or keep them out and lean on the in-house experience.

Separate lines for human and machine purchases

In a corporate blog post, Walmart explained that an agent shopping for a customer today signs in with that customer's own account, so human and machine purchases appear together with no distinction. Agent Profiles changes that: each agent's purchases show up on their own line, and shoppers can require sign-off on every order or allow the agent to buy within a spending limit they establish.

Walmart's examples set the bar at $50, $100, or $200. Approval is the default, and the retailer can request additional approvals even for purchases made within an agent's limit.

Walmart described the setup as analogous to adding an authorized user on a credit card: the agent gets a way to pay, the customer sets the limit, and the profile can be turned off at any time.

Walmart Agent Gateway

A profile is created automatically the first time a customer shops through an agent on the new Walmart Agent Gateway, and the customer connects it by signing into a Walmart account. Any Walmart account will do — no Walmart+ membership is required.

The gateway is a new server built on the Model Context Protocol (MCP), an open standard created by Anthropic that allows AI agents to connect to outside tools and data. Until now, agents have shopped Walmart through a website designed for people, which Walmart calls slow and error-prone. The gateway gives agents a direct route into the catalog and checkout and is launching in beta for businesses and developers. In practice, Walmart is treating agents as a distinct class of shopper with their own interface, rather than as traffic navigating pages built for humans.

It can also point an agent toward a customer's go-to items — sourdough bread or a specific brand of milk, for example — without exposing the person's entire purchase history.

Walmart is also dictating terms to agents. Payment-card data never reaches the agents, which cannot add or modify payment methods and can only access data pertinent to the task at hand. Agents that break the terms are banned from the Walmart site, and even when an agent does not act as intended, Walmart's standard returns process applies.

Furner pitched the plan at Masters of Scale in San Francisco

Walmart President and CEO John Furner shared the plan at the Masters of Scale Summit in San Francisco. The company has already signed on as a Muse shopping connector, Meta announced at its Connect event in September, as Cryptopolitan reported. Walmart also holds integrations with ChatGPT, Gemini, and others.

Neil Saunders, a retail analyst at GlobalData Retail, said Amazon is protecting a closed ecosystem and may be worried about agents eating into its advertising business. If outside agents play a much bigger role in discovery and purchasing, he added, the calculus could shift.

Consumer trust remains a sticking point: just 16% of shoppers say they feel comfortable with an AI assistant finding and purchasing products for them. In the meantime, Amazon has Alexa for Shopping, while Walmart's own assistant, Sparky, saw 60% growth quarter over quarter, Walmart U.S. CEO David Guggina said at a Goldman Sachs conference in September.

With Agent Profiles due next month and the gateway in beta, the weeks ahead will show whether retailer-supervised controls are enough to make shoppers comfortable delegating a cart to software.

Pricing dispute

Walmart is also facing accusations from Lindsay Owens, CEO of the Groundwork Collaborative, that it uses AI to alter prices based on shopper behavior. The company called the claims misleading and inaccurate.

"We price products, not people, so the prices that an AI agent sees are the same prices that a human would see on Walmart.com," Walmart said in its blog post.

Separately, Mastercard reportedly used an AI agent this week to pay two vendors nine times from one approved budget, according to Cryptopolitan. Its Agent Pay for Machines engine allows agents to pay for services as prescribed by a human.