Walmart opens its store to AI shopping agents with spending limits while Amazon shuts them out
Walmart announced on October 21 that it will let AI shopping agents buy on a customer's behalf through separate Agent Profiles, due within a month, with purchase approval by default or a spending cap the customer sets. Agents never see card details and cannot change payment methods. The Walmart Agent Gateway, an MCP server, is in beta for developers. Amazon blocked Meta's Muse agent in September.
This story was produced through MarketScale. See how Retail teams put it to work with Sales Enablement.
Key takeaways
Agent Profiles let customers set spending caps ($50, $100, $200) or require approval on every purchase, with agents unable to access card details or payment methods.
Walmart’s Agent Gateway runs on the Model Context Protocol, letting AI agents search, build carts and check out without browser automation.
Amazon blocked third-party agents citing lack of authorization to access customer accounts, scrape data, or process transactions.
Free workspace
Turn your Retail expertise into content.
Record interviews, organize footage, and write with AI on a free trial of the MarketScale platform for qualifying companies. No demo required, no credit card.
Walmart has decided that AI agents are customers. On October 21, at the Masters of Scale Summit in San Francisco, Walmart Inc. president and chief executive John Furner laid out a system for letting an AI assistant shop on a person's behalf, with its own profile, its own spending limit and its own set of rules, The Next Web reported. Amazon, a month earlier, had blocked Meta's Muse agent from shopping its site. The two largest retailers in the country have now taken opposite positions on the same question.
In September it blocked Meta's Muse, saying it had not authorized the agent to access customer accounts, scrape data or process transactions, Axios reported.
Agent Profiles for AI shopping
The consumer-facing piece is Agent Profiles, due within a month of the announcement. Each agent gets a separate profile inside the customer's Walmart account, so its purchases are kept apart from the customer's own. The default is that the customer approves every purchase. A customer can instead set a cap, with $50, $100 and $200 given as examples, and let the agent buy within it, though Walmart says it may still ask for extra approval inside the limit. Walmart compared it to adding an authorized cardholder to a credit card. Profiles can be deleted at any time, and a Walmart+ membership is not required, per The Next Web.
Security and control measures
The guardrails are the part retail operators should read closely. Agents never receive card details and cannot add or change a payment method. They may use only the information a task needs, must not act beyond what the shopper intends, and are blocked from the site if they break the terms. Walmart's normal returns process applies, including when an agent does something the shopper did not want. The company also said agents will see the same prices a human sees on Walmart.com, which matters because Furner had already ruled out personalized pricing in September after criticism from the Groundwork Collaborative, which Walmart called misleading.
Walmart Agent Gateway technology
Underneath it is the Walmart Agent Gateway, now in beta for companies and developers. It is a server built on the Model Context Protocol, the open standard Anthropic first published, which gives an agent a structured way to search, build a cart and check out instead of clicking through the website in a browser. Walmart says the gateway is faster and less error-prone than browser automation and can surface a customer's usual items without exposing their wider shopping history. Walmart already sells through Meta's Muse, ChatGPT, Gemini and other AI apps; it has not said which agents will connect through the gateway first.
Amazon's contrasting stance
Amazon's position is the opposite. In September it blocked Meta's Muse, saying it had not authorized the agent to access customer accounts, scrape data or process transactions, Axios reported. TipRanks reported Amazon has also kept other third-party shopping agents off its platform. Schwab and JPMorgan have taken a similar line with agents that try to move customers' money, per The Next Web. The logic on that side is control: an agent that shops anywhere is an agent that compares prices everywhere, and a platform built on being the default place to buy has little reason to make itself one option among many.
The broader implications for retailers
Walmart's bet is the reverse. If shoppers increasingly start in a chat window rather than a search box, the retailer that is easiest for an agent to buy from wins the order, and the one that blocks agents loses it without ever seeing the customer. Opening the gateway, with hard limits on money and data, is a way to be present in every assistant at once while keeping the account relationship.
For retailers and brands below the two giants, the decision is coming to them next. An MCP server is not an exotic build, and the governance Walmart described, separate agent profiles, default approval, spending caps, no card data exposure, is a reasonable template for anyone who wants agents as a channel without handing them the keys. The harder question is the one Amazon is asking: what an agent does with a store's catalog and prices once it has structured access to them.
Early surveys cited in the coverage put shopper comfort with letting an assistant find and buy products in the mid-teens as a percentage. — Walmart is building for the number it expects that to become.
Sources
- Walmart will let AI agents shop for you, with their own spending limits ↗ · The Next Web
- AI shopping wars: Walmart welcomes AI agents while Amazon blocks Muse ↗ · Axios
- Walmart Opens Its Store to AI Agents, With Spending Limits ↗ · The Keyword
- Walmart Agent Gateway: MCP server for AI shopping agents ↗ · TechAIWire
- Amazon blocks AI shopping agents from its e-commerce platform ↗ · TipRanks
Your experts belong here
Every story in MarketScale Retail starts with a company putting its merchandising leads, store operations teams, and category managers on the record. Buyers are already reading this topic. The only question is whose experts they find.
Category buyers trust operators, so your merchandising leads shorten the distance between first search and first call.