🔍 Read the full analysis: Amazon Blocks AI Shopping Agents, Walmart Lets Them In. Who Owns The Customer When An AI Buys? on ThorstenMeyerAI.com
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Amazon has blocked Meta’s Muse from checking its shopping site, while Walmart is opening an agent gateway and account profiles for authorized AI shopping tools. The split highlights a developing contest over how agents access stores, who controls checkout and customer data, and how transactions are attributed.
Amazon has blocked Meta’s Muse shopping agent from accessing its site, while Walmart is building an Agent Gateway for approved AI agents to search its inventory and make purchases. The competing approaches put retailers, technology companies and shoppers in a dispute over who controls the customer relationship when software handles shopping.
Meta’s Muse can browse websites and complete tasks on a user’s behalf. In a demonstration reported by The Verge, a reporter asked Muse where granola bars cost least per piece. The agent named Sam’s Club, Walmart and Costco, then said it could not check Amazon because Amazon had blocked it. The example shows a practical limit: an agent can compare only the stores it can access.
Amazon said Muse had not identified itself as an agent and appeared to capture and store customer credentials. The company said Meta had not told it the tool would shop on Amazon. Meta’s launch post, by contrast, said Muse had “no visibility into people’s passwords or payment methods.” Those statements describe opposing accounts of the agent’s access and handling of sensitive information.
On October 21, Walmart announced Agent Profiles and a Walmart Agent Gateway, which is based on the Model Context Protocol. The company’s approach is to provide approved agents with a structured way to search inventory and buy, rather than have them operate the ordinary website as a human would. Meta has also announced retail connections with Walmart and several other companies, including Best Buy, Gap, Sephora and Wayfair.
Agentic shopping
Three ways an AI agent buys for you
Every fight between shops and AI agents comes down to which of these three routes the agent uses, and whether the shop agreed to it.
The browser, like a human
The agent opens pages built for people, signs in with your login and clicks through checkout. Works almost anywhere, which is why shops object: they can't tell the agent from you.
Examples: Meta's Muse (from a virtual machine in Meta's cloud), Perplexity's Comet (from your own computer)
The shop's own agent door
Approved agents search and buy through a structured interface instead of clicking through pages. The shop sets the rules.
Examples: Walmart Agent Gateway (MCP, beta), Shopify MCP server and WebMCP checkout tools
Shared protocols
Common rules for who the agent is, what you allowed, and how the checkout and payment are confirmed.
Examples: Personal Agent Protocol (draft), Universal Commerce Protocol, Agentic Commerce Protocol
The protocols competing to be the standard
| Protocol | Backers | Covers | Status |
|---|---|---|---|
| Personal Agent Protocol (Poppy) | Sierra, Meta, Walmart, Shopify, Stripe + 35 design partners | Agent identity, customer permission, one session across site, API and company agent | Draft 0.1 (9 Oct 2026) |
| Universal Commerce Protocol (UCP) | Google, Shopify; site also lists Etsy, Wayfair, Target, Walmart, Amazon, Microsoft | Discovery, cart, checkout, orders; payment mandates via AP2 | Published |
| Agentic Commerce Protocol (ACP) | OpenAI, Stripe | Checkout between agent and merchant | Open source; ChatGPT's own checkout scaled back in March 2026 |
What a careful agent purchase looks like
- You give the task and a limit, for example a $50 cap.
- The agent identifies itself to the shop and gets only the access you approved.
- It prepares the cart and shows you an approval card.
- Payment goes through a token or one-time card, so the agent never sees your card.
- The shop’s normal returns apply if the agent got it wrong.
Sources: Meta (8 Sep 2026), Walmart (21 Oct 2026), TechCrunch on Shopify (28 Sep 2026), Sierra (9 Oct 2026), ucp.dev, agenticcommerce.dev, OpenAI (24 Mar 2026). Purchase steps: Walmart’s agent terms and Meta’s Muse description.
Powered by Thorsten Meyer AI
Retailers Set the Rules for Agents
The disagreement is about more than whether a bot can browse a product page. An agent can influence where a customer shops, which products it recommends, and which checkout and payment systems handle the transaction. A retailer that provides its own approved route can set conditions for access; a technology company that controls the agent may shape discovery and hold the shopper’s attention.
That creates consequences for businesses beyond the retailer and agent provider. Affiliate publishers may lose referral credit if an agent reads a review but completes a purchase without using a tracked link. The source report notes that Amazon Associates commissions depend on customers clicking tracked links; it does not establish how all agent purchases will be credited. For shoppers, agent purchases also raise questions about spending controls, payment details, and which company’s return rules apply.
Adoption figures cited by Meta suggest the potential audience is growing, but they are company-reported figures rather than an independent measure. On its October 21 earnings call, Meta said more than 10 million people in the United States and Canada had set up Muse within about six weeks, and described shopping as its most common use. Meta also said the average purchase was over $200. The figures indicate the company’s reported early use, not the overall size or reliability of AI shopping.
Products Worth Considering
Amazon Kindle Paperwhite Signature Edition (12th Generation, 2024), 32 GB, Metallic Black
Bose TV Speaker (2nd Gen) Compact Soundbar with Bluetooth and HDMI-ARC, Black
Amazon Kindle Paperwhite (12th Generation, 2024), 16 GB, 7-inch Display, Jade
As an affiliate, we earn on qualifying purchases.
From Browser Bots to Retail Gateways
Muse launched on September 8 as an agent able to use a browser, fill forms and act on a user’s behalf. It runs on a dedicated virtual machine in Meta’s cloud, according to the source material, and uses Link by Stripe for payments with a one-time-use card. Amazon restricted its access on September 20. At Meta Connect three days later, the company announced retail connections and Mark Zuckerberg said Meta expected to earn “a small fee from transactions.”
Retailers and technology firms are now pursuing more structured ways for agents to interact with commerce systems. Shopify introduced tools for browser agents to complete checkout through defined functions, while Walmart’s gateway is intended to provide an approved access route. Separate protocols are also being developed, including the Personal Agent Protocol, backed by Meta, Sierra, Walmart, Shopify and Stripe, and the Universal Commerce Protocol associated with Google and Shopify. Their functions and adoption differ, and the Personal Agent Protocol was still a draft in the source’s October 2026 account.
There is also a gap between shopping research and completed purchases. The source cites HUMAN Security data from September indicating that 81% of agent requests went to product and search pages, while 2.4% went to checkout and payment. Those figures describe the requests measured in that dataset and period; they do not establish how often all AI shopping agents make purchases.
“Continued access by an unauthorized AI agent violates Amazon’s Conditions of Use, to which our customers have agreed.”
— Amazon
Access, Safety and Sales Remain Unsettled
It is not clear how many Muse users attempted Amazon purchases, how often the agent has completed purchases across connected retailers, or how much revenue its transactions generate. Meta’s and Amazon’s accounts of Muse’s access to credentials also differ, and the source material does not independently resolve that dispute.
The emerging protocols have not established a single standard. Some are published, while others remain drafts or are supported by a limited group of partners. The source also describes earlier agent-shopping efforts that changed direction: OpenAI stepped back from its initial Instant Checkout approach in March 2026, saying it would let merchants use their own checkout experiences while focusing on product discovery. It remains unclear which models, if any, will gain broad adoption.
Retail access may also vary by company and over time. The source reports that Muse encountered Target’s bot check on one day and got through on another, and that users have had trouble with services including Delta and Amtrak. These accounts do not establish a uniform policy across those companies.
Watch for Wider Retail Adoption
The next test is whether Walmart’s Agent Gateway and Agent Profiles become available to customers and outside agent developers at meaningful scale. Retailers and protocol backers will need to clarify which agents are authorized, what permissions customers grant, and how orders, payments and returns are handled.
Shoppers considering agent purchases should check any available spending limits and purchase-approval settings, and understand which retailer processes the order. The source recommends using controls such as per-purchase approval, a spending cap or a one-time card; their availability depends on the agent and payment provider. Amazon’s access policy, Meta’s retail connections and the competing commerce protocols are likely to keep changing as companies negotiate who can participate in agent-led shopping.
Key Questions
Why did Amazon block Meta’s Muse?
Amazon said Muse was an unauthorized agent, did not identify itself and appeared to capture and store customer credentials. Meta said Muse could not see users’ passwords or payment methods. The source material does not independently settle the disagreement.
How is Walmart’s approach different?
Walmart announced an Agent Gateway and Agent Profiles for approved agents. The gateway is intended to let agents search inventory and buy through a structured route instead of interacting with the ordinary shopping site like a person.
Can AI agents complete purchases reliably?
The available figures suggest many measured agent requests are for product and search pages rather than checkout. HUMAN Security data cited in the source put 81% of agent requests on product and search pages and 2.4% on checkout and payment in its September dataset. Those figures do not describe every agent or every transaction.
What should shoppers check before using an agent?
Check whether purchases require approval, whether a spending cap or one-time payment card is available, and which retailer is responsible for payment, delivery and returns. Features vary by agent and merchant.
Source: ThorstenMeyerAI.com
