What Walmart announced

Walmart says it will let customers send their own AI agents to shop at Walmart and Sam's Club, with permissions and spending limits the customer sets. In an announcement on its corporate newsroom on 21 October 2026, the retailer introduced three pieces: Agent Profiles, a dedicated identity for each agent inside a customer's account; the Walmart Agent Gateway, a Model Context Protocol (MCP) interface built for agents rather than people; and a set of terms that spell out what agents may and may not do on its site.

Walmart president and CEO John Furner discussed the plan at the Masters of Scale Summit in San Francisco the same day, according to TechCrunch. Agent Profiles are due to launch within the next month, while the Gateway starts as a beta for companies and developers. Walmart frames the move as optional: "Shopping with an agent won't be for everyone, and it doesn't have to be," the company wrote, adding that its goal is to make its experience the most trusted for customers who do choose one.

The announcement builds on integrations Walmart already runs. The company lists Sparky, its own shopping assistant in the Walmart app, and conversational commerce integrations with Meta's Muse, ChatGPT, Gemini and others. What is new is a general set of rules and tools for any agent a customer picks, rather than a series of one-off partner deals.

Agent Profiles and spending limits

Walmart describes the problem Agent Profiles are meant to fix in plain terms. Today, when a customer uses an AI agent, the agent signs in with the customer's own account, so every purchase, whether made by the person or the software, shows up in the same order history with no way to tell them apart.

With Agent Profiles, each agent gets its own profile inside the customer's account. Purchases it makes are listed separately, and the customer chooses how approval works: either every purchase needs a yes, or the agent can buy on its own up to a spending limit. Walmart gives $50, $100 and $200 as example limits. A customer can turn an agent off, or delete its profile, at any time.

Walmart compares the setup to adding an authorized user to a credit card, with that person's spending identified and capped. The analogy is useful but has a limit of its own: an agent is not a family member, and its behaviour depends on the model and the company behind it. That is why the company pairs the profiles with rules for the agents themselves. Walmart also says prices do not change for agents. "We price products, not people," it wrote, so an agent sees the same prices a person would on Walmart.com, with applicable offers and fees shown before purchase.

Agent support does not require a paid membership. TechCrunch reports that it is available to anyone with a Walmart account, so a Walmart+ subscription is not needed.

The Agent Gateway runs on MCP

The second piece addresses how agents shop today. Many browse Walmart.com the way a person would, clicking through pages designed for human eyes and fingers. Walmart says that can be slow and error-prone, and that agents may get stuck along the way. The Walmart Agent Gateway is meant to replace that with a direct, structured connection over MCP, the open protocol Anthropic introduced for linking AI models to outside tools and data. TechCrunch describes the Gateway as Walmart's new MCP server, letting approved agents search Walmart's online inventory and make purchases for their users.

Walmart says the Gateway also lets it keep some data out of the agent's hands. Its example is a customer who usually buys sourdough bread or a particular brand of milk: Walmart will help the agent find those usual items without sharing the customer's broader shopping history with it. In other words, Walmart keeps the purchase history and answers the agent's question, instead of handing the agent the full record.

This puts Walmart on the server-side end of the agentic commerce spectrum. On the same day, Commerce.com and Cloudflare described turning on WebMCP across Commerce.com storefronts, which lets agents call store tools from inside the shopper's browser session. Shopify earlier opened its catalogue and Shop Pay checkout to AI agents through Shop MCP. Walmart's version sits closer to Shopify's: a dedicated server that agents connect to, with Walmart deciding what comes back.

The first time a customer shops through the Gateway with a given agent, Walmart automatically creates an Agent Profile for it. The customer then signs in to their Walmart account to authorize and link that profile, a step Walmart says only it sees, and sets a spending limit.

The rules agents must follow

The third piece is a set of terms for AI agents. Walmart lists them as expectations for how agents must operate when shopping on a customer's behalf:

Customers are in control, and Walmart's normal returns process applies, including when an agent does not act as intended. Customers must approve agent purchases by default, and Walmart may ask for extra approval even within an agent's spending limit. Agents must not take actions beyond what the shopper intends. They do not receive customers' payment card details and cannot add or change payment methods. They may access only the information needed for the task and allowed by the customer.

On privacy, the agent's own privacy policy covers what the agent receives, while Walmart handles the information an agent passes to it under Walmart's privacy notice. Agents that do not follow the terms will be blocked from Walmart's site.

The returns rule is the most concrete promise for shoppers. A misbehaving agent that orders the wrong item, or too many of something, is treated like any other purchase that can be sent back. As TechCrunch notes, that removes some of the risk of being an early user of agent shopping. The block for non-compliant agents is the enforcement side, though Walmart has not described how it will detect violations.

Why Walmart is going the other way from Amazon

Walmart's approach stands in contrast to Amazon's. The Verge reports that Amazon blocked Perplexity's shopping agent and Meta's Muse assistant, and TechCrunch notes that Amazon recently began blocking agents from making purchases on behalf of shoppers, while travel and booking sites have shut agents out entirely. The Verge summed up Walmart's position as letting agents in as long as they follow its rules.

Walmart is also one of the partners behind the Personal Agent Protocol that Meta and Sierra announced in early October, an open standard for how personal agents authenticate with businesses and what they are allowed to do. Walmart's announcement does not mention that protocol by name, but its building blocks of authorized sessions, scoped permissions and company-set limits point in the same direction.

The business logic is straightforward. If shoppers start handing routine orders to agents, a retailer that blocks them risks losing those orders, while one that welcomes them can try to shape how the agent shops: which data it sees, how it checks out, and what happens when it gets something wrong. Walmart is betting that clear controls will make customers comfortable enough to try.

How it will work for shoppers

Walmart describes three steps. First, start in the agent's app: the first time the agent shops through the Walmart Agent Gateway, Walmart creates an Agent Profile for it. Second, connect through Walmart by signing in and linking that profile to the account. Third, set controls: choose a spending limit, review purchases from every agent, and delete a profile at any time.

By default, every purchase needs approval, according to both Walmart and The Verge. Customers who want an agent to handle a routine grocery top-up without checking in can raise that to a budget, such as up to $100. The same tools apply at Sam's Club.

What customers will not get yet is a list of supported agents. Walmart has not named which assistants will connect through the Gateway at launch, and the Gateway itself begins as a beta for companies and developers, so the shopper-facing experience depends on which agent makers sign up.

What is still unclear

Several details are missing from the announcement. Walmart has not given a firm launch date beyond "the next month" for Agent Profiles, has not published developer documentation for the Gateway's MCP tools, and has not said how agent makers get approved or whether it charges them anything. It also has not explained how it will tell a compliant agent from a non-compliant one, or what happens to an agent that still browses the website the old way.

There are open questions about the model, too. An agent working through the Gateway relies on Walmart to decide which products to surface, which could shape what it recommends. Walmart's line that it prices products, not people, addresses one worry, but it does not say how search ranking or sponsored placement will be handled for agents. Prompt injection, where a malicious page or listing tries to steer an agent, is not addressed directly, although the approval defaults and spending caps limit how much damage a hijacked agent could do.

Finally, the rules are Walmart's own. They protect customers on one retailer's site, but a shopper using an agent across several stores will still face different terms at each, which is the gap shared standards are trying to close.

Common questions

When can I use an AI agent to shop at Walmart?

Walmart says Agent Profiles launch within the next month. The Agent Gateway starts as a beta for companies and developers, so it depends on which agents connect to it.

Can an agent spend my money without asking?

Not by default. Every agent purchase needs approval unless you set a spending limit, such as $50, $100 or $200, and Walmart may still ask for extra approval.

Do I need Walmart+?

No. TechCrunch reports that agent support is available to anyone with a Walmart account.

Sources & further reading

  1. Walmart Makes It Possible for Agents to Shop on Behalf of Customers, While Keeping Customers in Control ↗
  2. While Amazon blocks AI agents, Walmart rolls out the welcome mat ↗
  3. Walmart is inviting AI agents to shop for you ↗
How this story was made

Written by Kristian Kostov with AI assistance and checked against the linked sources. Company performance claims are attributed to the company. Analysis reflects AiLookout’s interpretation; we have not independently tested the products discussed. The cover places the companies’ own logos, taken from public-domain or openly licensed files credited in the caption, on a plain background. It is not a product photograph, a screenshot, or an endorsement.

Our editorial standards
Back to all stories