Walmart AI shopping agents just went from an experiment to official company policy. On October 21, 2026, Walmart announced two connected tools that let AI assistants shop and complete purchases on its behalf of customers: Agent Profiles, which give every AI agent its own permissioned profile inside your Walmart account, and the Walmart Agent Gateway, a new platform that lets assistants like ChatGPT, Google Gemini, and Meta’s Muse connect directly to Walmart and Sam’s Club systems instead of fumbling through a web browser like a confused human.
Why does this matter to you? Because it is the clearest sign yet that the biggest players in retail are splitting into two camps. Amazon is actively blocking AI agents from buying for shoppers. Walmart just rolled out the welcome mat for Walmart AI shopping agents, built the infrastructure, and wrote the rulebook. If you have ever wanted to say “buy my weekly groceries” and have it actually happen, this is the announcement that moves that fantasy toward reality, with guardrails that keep you in control.
This guide breaks down exactly what Walmart announced, how Agent Profiles and the Agent Gateway work, the safety rules Walmart wrote for Walmart AI shopping agents, and what it all means for your wallet. As of publication, Agent Profiles are rolling out next month (November 2026), and the Gateway is in an invitation-only beta, so this is what is coming, not what is live today.

What Walmart Actually Announced
On Wednesday, October 21, 2026, Walmart introduced a package of tools designed to let Walmart AI shopping agents search products and complete purchases for customers, according to reporting by Reuters. The announcement has three parts that work together:
- Agent Profiles — dedicated profiles inside your Walmart account for each Walmart AI shopping agent you authorize. Each profile tracks that agent’s purchases separately from your own and carries the spending rules you set. Rolling out next month.
- The Walmart Agent Gateway — a platform that connects AI assistants (OpenAI’s ChatGPT, Google’s Gemini, and Meta’s Muse are the named partners) directly to Walmart and Sam’s Club systems, so Walmart AI shopping agents get real product and fulfillment data instead of scraping the website.
- Terms for AI Agents — a formal rulebook governing what agents can and cannot do on Walmart’s sites, including who approves purchases and what happens when an agent misbehaves.
Walmart CEO John Furner unveiled the move at the Masters of Scale Summit in San Francisco, framing it as infrastructure for a future where Walmart AI shopping agents shop at scale across both Walmart and Sam’s Club. The reaction from Wall Street was notably warm: Mizuho analyst David Bellinger wrote in a note that the tools instilled confidence in what he called the “experimental” shopping process of agentic purchases, adding that “the momentum and buy-in from top management teams seems to be making this into a real thing.”
Before this announcement, Walmart AI shopping agents were already possible in a messy way: an agent would sign in using your actual account and click around the site like a person. Walmart described that approach as slow and error-prone. The new system replaces the mess with structure — and that structure is where the interesting part lives.
Agent Profiles: A Permission Slip for Every AI
Think of an Agent Profile the way you would think of a family member on your account, but with stricter rules. When you authorize Walmart AI shopping agents to shop for you, Walmart creates a dedicated profile for that specific agent inside your account. Your agent gets its own lane, its own purchase history, and its own limits. You never have to hand over your main account credentials again.
Spending limits and per-transaction approval
The headline feature of Agent Profiles is control. For each Walmart AI shopping agent, you choose one of two modes:
- Require approval for every purchase. Nothing gets bought until you say yes to that specific transaction. Maximum caution, maximum involvement.
- Set a spending limit. The agent can buy freely within the cap you define, and Walmart gave $50, $100, and $200 as examples of the limits you might choose. A $50 cap for your weekly grocery agent and a $200 cap for a household-supplies agent is the kind of setup this makes natural.
Two important details from Walmart’s terms keep the power with you even after you set a limit. First, customers approve agent purchases by default — the system starts cautious and loosens only where you allow it. Second, Walmart says it may require additional approvals even for purchases that fall within an agent’s spending limit, so the retailer keeps a backstop for edge cases. Permissions can be switched off at any time, and the whole thing is available to every Walmart account holder — no Walmart+ membership required.
Walmart’s own analogy is telling: the company compared an Agent Profile to adding someone as an authorized cardholder on your credit card, with the agent getting its own profile to shop under limits you control. That framing is deliberate. It signals that Walmart wants shoppers to think of Walmart AI shopping agents not as anonymous bots but as named, permissioned helpers with visible boundaries.

Agent purchases tracked separately
One of the most practical features is also one of the simplest: purchases your agent makes show up separately from your own order history. Under the old, messy setup, an agent signed in as you, so its purchases and yours were jumbled together with no differentiation. If the agent bought the wrong brand of paper towels, good luck spotting it in a list of forty orders.
With Agent Profiles, every Walmart AI shopping agent has its own visible trail. You can view purchases from every agent you have authorized, compare them against your limits, and spot mistakes quickly. That separation is also what makes revoking access meaningful — when you can see exactly what one agent did, you can decide whether it keeps its privileges.
How the setup flow works
Walmart described a setup that starts in the Walmart AI shopping agent’s own app, not on Walmart’s site. The first time you use an AI agent to shop through the Walmart Agent Gateway, Walmart automatically creates an Agent Profile for that agent. You then authorize it, link it to your Walmart account, and choose your spending limit. The profile can be deleted at any time, which instantly cuts that agent off.
A few promises for Walmart AI shopping agents are baked into the design. Walmart said its Every Day Low Price commitment is unchanged whether you shop directly or through an agent: the prices an AI agent sees are the same prices a human would see on Walmart.com, with applicable offers and fees shown before purchase. That matters more than it sounds — one of the quiet fears about agent shopping is that retailers might show bots worse prices or hide fees in the automated flow. Walmart has now put in writing that this will not happen on its platform.
The Walmart Agent Gateway: A Faster Lane for Agents
If Agent Profiles are the permission slips, the Walmart Agent Gateway is the road they drive on. Today, Walmart AI shopping agents have to do what you do: load pages in a browser, click buttons, scroll, and fill forms. Walmart called that process slow and error-prone, and it is easy to see why. Web pages are built for human eyes and human reflexes; every layout change can break an agent’s script, and a single misclicked button can order twelve jars of pickles instead of one.
The Gateway gives Walmart AI shopping agents a direct pipe into Walmart’s systems. Instead of scraping the website, authorized Walmart AI shopping agents receive up-to-date product and fulfillment data straight from Walmart — stock levels, prices, delivery windows — with what the company describes as reduced requests, lower latency, and lower token usage. That last point is a quiet nod to developers: fewer wasted tokens means cheaper Walmart AI shopping agent runs, which means agent shopping stays affordable as a service.
Why Walmart built it on MCP
The technical foundation is the Model Context Protocol (MCP), the open standard originally created by Anthropic, which lets AI assistants plug into external tools and data sources in a consistent way. By building the Agent Gateway as an MCP server, Walmart is betting on an open protocol rather than a proprietary API that locks one assistant in and everyone else out. In practice, that means any Walmart AI shopping agent platform that speaks MCP — and most now do — can integrate with Walmart through the same door.
The Gateway also solves a subtle privacy problem. One of its advertised abilities is helping Walmart AI shopping agents shop for your usual items — your sourdough bread, your particular brand of milk — without sharing your personal information or broader purchase history with the agent. Think about what that means: the agent does not need to know everything you have ever bought to know what you buy every week. Walmart hands the agent the shopping pattern it needs and nothing more. That is a meaningful design choice, and it directly answers the privacy anxiety that has shadowed the whole agentic-commerce conversation.

Walmart’s official beta page for the Gateway lists the safeguards built into the system: verified identities for agents, controlled permissions, observable actions, plus category restrictions, rate limits, and compliance guardrails. “Observable actions” is the phrase to notice — it means what the agent does through the Gateway is logged and visible, not a black box.
Who gets in (and when)
Here is the honest timeline. The Gateway is being built out now, beginning with a beta that lets companies and developers build Walmart AI shopping agents. Walmart’s beta page is explicit that this is an early program: access is limited, available by invitation only, may be suspended at any time, and some features may change and may not be available to all participants. In other words, developers are getting the keys first, consumers get the polished version later, and Walmart reserves the right to adjust the rules as it learns.
Agent Profiles, by contrast, are the consumer-facing piece, and they roll out next month. The two launches are sequenced deliberately: the Gateway gives agent makers the infrastructure, and Agent Profiles give shoppers the controls. One without the other would be either powerful but dangerous or safe but useless.
The Safety Rules: Walmart’s Terms for AI Agents
Buried under the product announcements is the part that may matter most in the long run: Walmart wrote a formal rulebook for Walmart AI shopping agents and published it in its site Terms of Use (last updated September 21, 2026). This is one of the first times a major US retailer has codified, in plain legal language, exactly how autonomous software is allowed to behave on its platform. Other retailers are watching — and so should you, because these terms decide where the liability lands when an agent gets something wrong.
Agents never touch your card data
The single most important protection: Walmart AI shopping agents are barred from accessing your payment card data. They cannot add a new payment method, and they cannot change the payment methods already on your account. When one of your Walmart AI shopping agents buys something, it triggers a charge through the payment rails you already set up — it never sees the card number, the expiry date, or the CVV. Walmart’s terms state that agents can access only the information needed for the task at hand, and only what the customer allows.
That separation is the difference between delegating a task and handing over your wallet. A rogue or buggy Walmart AI shopping agent cannot skim your card, cannot open a new line of payment, and cannot quietly switch your checkout to a different card. For shoppers who have been nervous about letting software spend their money, this is the guardrail that matters most.
Walmart also applies its normal returns process to Walmart AI shopping agent purchases — including cases where a Walmart AI shopping agent does not act as intended. If your agent orders the wrong item, the return works the same way as if you had clicked “buy” yourself. And critically, the terms state that actions authorized Walmart AI shopping agents take within their granted permissions are treated as the customer’s own. You authorized it; its purchases are your purchases. That clarity cuts both ways: it protects Walmart’s checkout from disputes, and it tells you exactly where you stand.
How agents must behave
Walmart’s definition is broad: an “agent” is any software-based system, application, or service that autonomously or semi-autonomously accesses or interacts with Walmart’s sites on a user’s behalf or at a user’s instruction. That covers AI assistants and browser-automation tools alike. (Assistive technologies that help users with disabilities access the sites are explicitly excluded — an important carve-out.)
The behavioral rules for authorized Walmart AI shopping agents include:
- Identify themselves. Agents must not conceal that their interactions are automated. No pretending to be a human shopper.
- Stay inside the task. Agents must not take actions beyond what the shopper intends. An agent told to buy groceries should not be browsing electronics.
- Respect the limits. Agents must comply with rate limits and other technical measures, and Walmart may limit the value, quantity, or frequency of agent transactions — and can decline or reverse transactions outright.
- Accept suspension. Walmart can suspend or revoke an agent’s access at any time, and agents that violate the terms get blocked from the site entirely. Customers can also revoke an agent’s access themselves through account settings.
- No training on Walmart’s data. Using site materials to develop, train, or improve AI or machine-learning models without written consent is separately prohibited.
The terms also draw a clear line on responsibility: authorizing an agent does not make it a Walmart employee, partner, contractor, or representative, and Walmart states it is not responsible for a third-party agent’s conduct. Privacy-wise, the agent provider’s own privacy policy governs whatever information the agent receives, while Walmart processes what the agent provides under Walmart’s Privacy Notice. Read your agent provider’s privacy policy before you hand it shopping duties — Walmart is telling you, in so many words, that your data’s journey does not end at its door.

Why Walmart Said Yes While Amazon Said No
You cannot understand the Walmart announcement without the contrast that made it news. While Walmart was building infrastructure to welcome Walmart AI shopping agents, Amazon has been moving in the opposite direction — actively limiting third-party AI shopping agents’ access to its platform. Shortly after Meta launched its Muse personal agent in September 2026, Amazon blocked Muse from shopping on Amazon.com. The two largest US retailers have now placed opposite bets on the same technology.
This split is part of a wider standoff. Just days before Walmart’s announcement, websites blocking AI agents became one of the month’s biggest tech stories: Yelp requires agents to pay for data licensing, Delta Air Lines refuses to let agents book tickets, and JPMorgan Chase and Charles Schwab have closed account access to agents over security and liability concerns. Banks have separately warned about the risks of using AI agents for online shopping, including scams, fraud, and data-privacy breaches. The default posture of the web in late 2026 has been defensive — lock the agents out, figure out the rules later.
Walmart’s bet on Walmart AI shopping agents is that the rules can be figured out now, and that doing so is a competitive advantage. The retailer already works with Meta’s Muse and plugs into AI apps like ChatGPT and Gemini through conversational commerce integrations, and it runs its own Sparky shopping assistant. Rather than waiting to see what shopping becomes, Walmart said it wants to help shape it. If agentic commerce does become “a real thing,” as the Mizuho analyst put it, the retailer that built the on-ramps first owns the traffic.
There is also a defensive logic hiding inside the welcoming gesture toward Walmart AI shopping agents. Walmart AI shopping agents that shop through the official Gateway are ones Walmart can see, rate-limit, and revoke. An agent scraping the site with stolen session cookies is none of those things. By building the authorized path, Walmart is quietly shrinking the unauthorized one — the same reason it wrote unauthorized scraping and ordering into its prohibited conduct. Welcoming agents on Walmart’s terms is also a way of keeping them under Walmart’s watch.
What Walmart AI Shopping Agents Mean for Shoppers Like You
Strip away the strategy, and the consumer question is simple: what changes in your kitchen, your budget, and your weekly routine? Here is the practical picture of Walmart AI shopping agents, based on what Walmart has actually announced.
Weekly reorders become a sentence. The canonical use case for Walmart AI shopping agents is the boring, repetitive shop: the groceries you buy every week, the household supplies you always forget until you run out. “Reorder my usual groceries” is the kind of instruction the Gateway is designed for — your sourdough bread, your brand of milk, handled by Walmart AI shopping agents that know the pattern but not your life story. Walmart explicitly built the system so agents can handle these usual items without seeing your personal information or full purchase history.
Budgets become enforceable. A spending limit is not a suggestion in this system; it is a profile-level rule. Parents can give a Walmart AI shopping agent a $100 weekly cap. A small business can let a Walmart AI shopping agent for office supplies spend $200 a month and require approval above that. The per-transaction approval mode exists for anyone who wants to watch every single purchase. For households that treat budgeting as a discipline, this is automation with a leash.
Mistakes become traceable. Because your Walmart AI shopping agents’ purchases sit in their own history, auditing is straightforward. Did the agent substitute a brand you hate? You will see it in that agent’s trail, not buried in your own orders. And because Walmart’s normal returns process covers agent purchases, the fix for a wrong item is the same return flow you already know.
Prices stay honest. Walmart’s written commitment that agents see the same Every Day Low Prices as human shoppers — with offers and fees shown before purchase — removes one of the sneakier risks of agent commerce: the fear that the machine gets a worse deal than you would. Watch whether other retailers match this promise; it is the kind of standard that only matters once someone writes it down.

The catches are real, too. Agent Profiles do not exist yet — they arrive next month. The Gateway is in an invitation-only developer beta, so the polished consumer experience is still being built. And the liability line is unambiguous: an authorized agent’s purchases are your purchases, and Walmart disclaims responsibility for what a third-party agent does. If a Walmart AI shopping agent has a bad day and buys the expensive olive oil, the charge is yours and the return is on you to initiate. Convenience always has a price; here it is vigilance.
How to Get Started with Walmart AI Shopping Agents (When It Rolls Out)
There is nothing to sign up for today, but the path is clear from Walmart’s announcement. When Agent Profiles launch next month, expect the flow to start inside your chosen Walmart AI shopping agent’s app — ChatGPT, Gemini, or Muse. The first time you ask that assistant to shop through the Walmart Agent Gateway, Walmart will automatically create an Agent Profile for it. You then link the profile to your Walmart account, pick a spending limit (or require approval on every purchase), and you are set. From there, each agent’s purchases appear in their own section of your account, and you can adjust limits or delete profiles whenever you want.
For developers and companies building Walmart AI shopping agents, the Gateway beta is the door to knock on — access is invitation-only for now. If you build agent products, the reduced token usage and direct fulfillment data are the practical reasons to integrate early; the verified-identity and observable-action features are the reasons Walmart will let you.
One piece of advice worth taking seriously: read your agent provider’s privacy policy before you authorize anything. Walmart’s terms are explicit that the agent provider’s policy governs the information the agent receives. Your shopping data will touch at least two companies’ systems. Know what each one promises.
The Bigger Picture: Agentic Commerce in 2026
Walmart is not acting in a vacuum. The last month has been a pile-up of agentic-commerce milestones, and the pattern is unmistakable: money is learning to move through AI agents, and every layer of the stack is being rebuilt to allow it.
Just this week, Claude Duet arrived as an agent that works before you ask and can complete purchases through payment rails like Stripe Link and Shop Pay. Days earlier, Mastercard Agent Pay laid down the payment-network layer for agent transactions. Walmart’s announcement is the retail layer for Walmart AI shopping agents clicking into place: the store itself, rebuilt so agents can walk in through the front door instead of climbing through the window.
Put the three together and the shape of 2026’s agentic commerce stack emerges: assistants that decide (ChatGPT, Gemini, Muse, Duet), payment networks that settle (Mastercard Agent Pay, Stripe Link, Shop Pay), and merchants that fulfill (Walmart, via the Agent Gateway). Each layer is adding its own guardrails for Walmart AI shopping agents — spending limits here, verified identities there, card-data firewalls everywhere. The industry is converging on the same insight: agents can spend money only inside cages built by humans.
The open question is which cage design wins. Amazon’s answer is the wall: keep Walmart AI shopping agents out, build your own AI shopping tools (Alexa for Shopping, Buy for Me) instead. Walmart’s answer is the turnstile: let Walmart AI shopping agents in, but only through the Gateway, only with a Profile, only under limits, and only while behaving. Yelp, Delta, and the banks mostly chose versions of the wall. Over the next year, shoppers will vote with their habits — and the retailer whose approach feels both convenient and safe will set the template everyone else copies.
Frequently Asked Questions
When will Walmart Agent Profiles be available?
Walmart said Agent Profiles will roll out next month — November 2026. They are not live yet. The Walmart Agent Gateway is further along but currently in an invitation-only beta for companies and developers, so the full consumer experience is still being built out.
Which AI assistants can act as Walmart AI shopping agents through the Gateway?
Walmart named OpenAI’s ChatGPT, Google’s Gemini, and Meta’s Muse as the Walmart AI shopping agents connecting through the Gateway, and the platform covers both Walmart and Sam’s Club. Because the Gateway is built on the open Model Context Protocol (MCP), other MCP-compatible agents can potentially integrate as the beta expands.
Can Walmart AI shopping agents see my credit card number?
No. Walmart’s terms explicitly bar agents from accessing payment card data, and agents cannot add or change payment methods on your account. Purchases are charged through the payment methods you already set up — the agent triggers the transaction but never sees the card details.
Do I need a Walmart+ membership to use Walmart AI shopping agents?
No. Walmart said the tools will be available to all Walmart account holders and will not require a Walmart+ membership.
What happens if my Walmart AI shopping agent buys the wrong thing?
Walmart’s standard returns process applies to agent purchases, including when an agent does not act as intended. Because each agent’s purchases are tracked in their own separate history, you can spot the mistake quickly, return the item the normal way, and adjust or revoke that agent’s permissions.
Will a Walmart AI shopping agent get different prices than I would?
According to Walmart, no. The company said its Every Day Low Price commitment is unchanged whether you shop directly or through an agent — agents see the same prices a human would see on Walmart.com, with applicable offers and fees shown before purchase.
The Bottom Line
The arrival of Walmart AI shopping agents is not really a story about a retailer adding a feature. It is a story about the web’s biggest question in 2026 getting its first serious answer: when software starts spending your money, who sets the rules? Walmart’s answer is detailed, written down, and already shipping — profiles with spending limits, a Gateway with verified identities, and terms that keep your card data behind a wall the agent cannot cross.
Amazon’s opposite bet means shoppers are about to run a live experiment whether they signed up or not. One camp says Walmart AI shopping agents are guests to be managed; the other says they are intruders to be blocked. Whichever camp makes “reorder my groceries” feel effortless and safe will define how the next decade of shopping works. For now, set a reminder for November: that is when Agent Profiles go live, and when you can finally tell an AI to do the weekly shop — with a $100 leash attached.
Sources: Reuters reporting on the October 21, 2026 announcement (via Indian Express and The Hindu BusinessLine); Unite.AI’s breakdown of the announcement; Walmart’s official Agent Gateway beta materials as described in that coverage; Technopolis News on the MCP-based Gateway design. Pricing examples ($50/$100/$200) and rollout timing are Walmart’s stated figures; Gateway availability remains invitation-only beta and features may change.
