Walmart welcomes AI shopping agents with spending controls
Walmart is officially opening its site to AI agents that shop on behalf of customers, in the same week Amazon started blocking them.
For you: You can hand shopping to an AI agent while you set the spending cap and decide whether each payment needs approval.

Online shopping is slowly moving toward autopilot. Instead of a person browsing, an AI agent can search for products and buy them on that person's behalf. Big retailers disagree on how to handle this. This week Amazon began blocking AI agents from buying things for shoppers on its site. Walmart went the other way and rolled out the welcome mat.
CEO John Furner announced the move on October 21 at the Masters of Scale Summit in San Francisco. The technology covers both Walmart.com and Sam's Club, the company's warehouse retailer.
What Agent Gateway is
Walmart launched Agent Gateway, a new MCP server. MCP, the Model Context Protocol, lets AI agents connect to outside data sources, tools and workflows. Anthropic originally created it.
Through the Gateway, an approved agent can:
- search Walmart's online inventory;
- make purchases for its user;
- shop for the brands and items you usually buy. Your personal information and purchase history are not shared with the agent.
Walmart had already announced a partnership with Meta's Muse agent, and it offers integrations with ChatGPT, Gemini and other AI apps.
Agent Profiles: keeping spending in check
Next month, in November, Walmart will roll out Agent Profiles. Each AI agent you use gets its own profile inside your account, so its purchases show up separately from yours.
You choose how purchases are approved:
- Approve every transaction. The agent asks before each purchase.
- Pre-authorize up to a limit. The agent can buy on its own up to a spending cap you set.
Walmart compares this to adding an authorized user to a credit card. That person's charges are clearly marked, and you can cap what they spend.
A profile is created the first time you use an agent to shop through the Gateway. You then link it to your Walmart account and set controls or limits. You can delete a profile at any time. Any Walmart account holder can use it, and a Walmart+ membership is not required.
Rules for the agents
Agents must follow special terms:
- they cannot take actions beyond what the shopper intends;
- they cannot access customers' payment information;
- they can only see the information needed for the task at hand;
- agents that break the rules get blocked from the site.
If an agent still gets something wrong, Walmart's standard returns process applies. That may reassure shoppers who are wary of letting AI buy things for them.
Why Amazon is taking a different path
Amazon, along with a number of travel and booking sites, has locked AI agents out entirely. The logic is easy to see: a store benefits when you visit in person and see its ads and recommendations. An agent skips all of that. Walmart chose to set rules for agents and treat them as a new sales channel instead. Which approach wins remains to be seen.
What this means for readers in Kazakhstan
Walmart does not operate in Kazakhstan, but the trend will reach local shoppers too. If you order goods from abroad, comparing prices and checking out through an assistant will get easier. Local marketplaces will also have to decide sooner or later whether to welcome or block agents.
What you can do now:
- When you let an agent shop for you, start with approval for every payment. Switch to a spending limit only once you trust it.
- Keep the limit low, for example close to what you spend on household basics in a month.
- Review the agent's purchase history once a week.
- Write clear instructions:
I need [product name]. The price must not exceed [amount] tenge. Compare several options and make a table with price, delivery time and return policy. Ask for my permission before buying anything.
That way you save time without overspending.
Photo: Shopping online with bank card, Bogdan Hoyaux / European Commission — CC BY 4.0
Source: techcrunch.com




Comments
Sign in to leave a commentNo comments yet. Be the first.