🔍 Read the full analysis: Retailers Face A New Question As AI Agents Shop For Customers on ThorstenMeyerAI.com
Get privacy and security gear 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
Meta says more than 10 million people in the United States and Canada set up its Muse shopping agent within about six weeks, though most agent activity documented in separate security data has been product research rather than checkout. Walmart is opening a structured route for approved agents, while Amazon has blocked Muse, sharpening questions about customer control, retailer access and how sales are attributed.
Walmart is opening a channel for AI shopping agents, while Amazon has blocked Meta’s Muse agent, setting up a dispute over which automated tools retailers will admit and how they will handle purchases. The competing approaches gained attention on October 21, when Walmart announced its Agent Gateway and Meta reported that more than 10 million people in the United States and Canada had set up Muse within about six weeks, according to figures the company gave on its third-quarter call.
Walmart’s Agent Gateway uses the Model Context Protocol, or MCP, to let approved agents search inventory and make purchases through a structured connection rather than by navigating a standard retail website. Walmart also announced Agent Profiles, which can set permissions and spending limits for an agent acting for a customer. The company’s announcement establishes the planned access model; details about availability and how widely it will be adopted remain limited in the source material.
Meta says shopping is the most common use of Muse and that its average purchase is worth more than $200. Those are Meta’s figures, not an independent measurement of completed transactions. A Verge reporter’s price comparison illustrates the access problem: Muse recommended Sam’s Club, Walmart and Costco for granola bars, but said it could not check Amazon because the retailer had blocked it.
Amazon’s stated objection is that Muse is an unauthorized agent that does not identify itself and appears to capture and store customer credentials. Meta’s launch post says Muse “has no visibility into people’s passwords or payment methods.” The accounts conflict, and the material provided does not independently resolve how Muse handles credentials or what Amazon observed. The dispute is about more than whether a bot can browse: it touches security, retailer control of customer accounts and the path by which a purchase is completed.
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 Terms of Agent Shopping
The stakes extend beyond convenience. If a customer delegates product research and purchasing to an agent, the retailer may lose direct control over how that shopper discovers products, compares prices and reaches checkout. Walmart’s gateway offers a way to participate while setting conditions; Amazon’s block keeps an outside agent from using its site through the route Muse attempted.
The commercial effects are still uncertain, but sales attribution and advertising are clear concerns. The source reports that Amazon earned more than $68 billion from advertising last year, citing GeekWire, and says Meta expects to earn “a percent of the value” Muse creates. If agents make choices without a conventional tracked referral, publishers and affiliate businesses may not receive credit under systems built around clicks. That is a potential consequence, not evidence that agent shopping has already displaced those sales.
Early usage claims also need to be read alongside activity data. HUMAN Security’s September figures, as presented in the source, say 81% of agent requests went to product and search pages, while 2.4% went to checkout and payment. Those percentages describe the dataset and time period reported by the company; they do not establish the rate for all shopping agents or prove that most consumers are already buying through them.
Products Worth Considering
Amazon Kindle Paperwhite (12th Generation, 2024), 16 GB, 7-inch Display, Jade
Bose SoundLink Flex Bluetooth Speaker (2nd Gen), Portable Waterproof Speaker, Black
Bose TV Speaker (2nd Gen) Compact Soundbar with Bluetooth and HDMI Connectivity
As an affiliate, we earn on qualifying purchases.
A Short, Fast-Moving Protocol Race
The dispute followed a sequence of announcements. Meta launched Muse on September 8, describing an agent that can open a browser and fill out forms. The source says Muse runs in a dedicated virtual machine in Meta’s cloud and pays through Link by Stripe, which issues a one-time-use card. On September 20, Amazon cut off Muse. Three days later, Meta announced retail connections involving Best Buy, Gap, Sephora, Walmart, Wayfair and Expedia.
Other companies are trying to make agent transactions more structured. Shopify introduced tools for browser agents to complete checkout, while a draft Personal Agent Protocol from Sierra and Meta was announced with Walmart, Shopify and Stripe and later attracted additional design partners. A separate Universal Commerce Protocol backed by Google and Shopify addresses the shopping journey, while OpenAI and Stripe’s Agentic Commerce Protocol focuses on checkout. These efforts are at different stages and should not be treated as one agreed industry standard.
Earlier agent-shopping products have also faced limits. The source says OpenAI stepped back from its initial ChatGPT Instant Checkout approach in March 2026, saying the first version did not provide the flexibility it wanted and that it would let merchants use their own checkout experiences while focusing on product discovery. That history suggests that building a connection is not the same as persuading retailers to adopt it or making customers comfortable with an agent completing a purchase.
““Continued access by an unauthorized AI agent violates Amazon’s Conditions of Use, to which our customers have agreed.””
— Amazon, in a message shown to Muse users
Usage, Security and Sales Remain Unsettled
Meta’s user and purchase figures are company-reported; the source does not provide an independent audit, a precise definition of “set up,” or enough detail to determine how many users completed purchases. The HUMAN Security figures cover a particular September dataset, and the source does not give enough information here to generalize them to all agents or retailers.
It is also unclear how Walmart will approve agents, what information each agent will receive, or how spending limits and customer permissions will work in practice. Amazon and Meta disagree over Muse’s identification and handling of credentials. The source offers their statements but no independent technical review that settles the dispute.
Finally, the commercial terms are unsettled: the material does not specify how agent referrals, advertising impressions, affiliate commissions or returns would be credited across the different systems. It also does not establish whether shoppers will consistently trust agents to choose products and complete payments on their behalf.
Gateway Rollouts and Protocol Decisions
The next test is whether Walmart’s announced gateway and agent profiles become available at meaningful scale and whether outside agents can use them under clear customer permissions. Retailers, payment companies and agent developers will also need to show how checkout, spending controls, returns and purchase records work across systems.
For consumers, the practical question is not just whether an agent can find a price, but who authorizes the payment and handles the order. Until the rules are clearer, users should check purchase approvals, spending limits, payment methods and which retailer’s return policy applies. The source describes a fast-moving market, but it does not establish a timetable for a common standard or broad retailer participation.
Key Questions
What did Walmart announce?
Walmart announced an Agent Gateway using MCP and customer Agent Profiles with permissions and spending limits. The announcement describes a route for approved agents to search and buy through a structured connection.
Why did Amazon block Meta’s Muse?
Amazon said Muse was an unauthorized agent that did not identify itself and appeared to capture and store customer credentials. Meta said Muse has no visibility into passwords or payment methods; the material provided does not independently resolve the dispute.
Are AI agents already completing most online purchases?
The cited September HUMAN Security data says 2.4% of agent requests went to checkout and payment, compared with 81% to product and search pages. That dataset does not establish a rate for every agent or retailer, and Meta’s separate usage figures are company-reported.
What could agent shopping mean for affiliate publishers?
Affiliate systems often depend on a customer clicking a tracked link. If an agent researches a product and completes a purchase without that click, a publisher may not receive the usual referral credit. The scale of that effect is not established.
Source: ThorstenMeyerAI.com
