🔍 Read the full analysis: When An AI Shops, Who Builds The Customer Relationship? on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Meta says more than 10 million people in the US and Canada set up its Muse agent in about six weeks, with shopping its most common use. Walmart is opening a structured route for agents to shop, while Amazon has blocked Muse, putting control of the customer relationship, checkout and sales attribution at issue.
Walmart announced tools for AI shopping agents on October 21, the same day Meta said more than 10 million people in the US and Canada had set up its Muse agent in about six weeks. The competing moves show retailers deciding whether to let outside agents search and buy through approved systems or block them, a choice that could shift control of shopping data, checkout and customer relationships.
Meta said on its third-quarter call that shopping is Muse’s most common use and that the average purchase made through the agent is over $200. Those figures are Meta’s own, and the company did not provide independent verification in the source material. Muse can use a browser to search and make purchases on a user’s behalf. In a September example reported by The Verge, it compared granola-bar prices at Sam’s Club, Walmart and Costco, but said it could not check Amazon because the retailer had blocked access.
Walmart’s Agent Gateway is designed to let approved agents search inventory and make purchases through a structured connection based on the Model Context Protocol. Walmart also announced Agent Profiles, which can set limits on what an agent is allowed to do. Amazon has taken the opposite position with Muse: it blocked the agent and said it had not been told Muse would shop on Amazon, did not identify itself as an agent and appeared to capture and store customer credentials. Meta has disputed that last point, saying Muse has no visibility into users’ passwords or payment methods.
The dispute extends beyond those two retailers. Shopify has introduced structured tools for browser agents to complete checkout, while Meta and Sierra published a draft of the Personal Agent Protocol, also called Poppy, with Walmart, Shopify, Stripe and other design partners. Separate commerce standards are also being developed. Their presence shows that companies are trying to set rules for agent identity and shopping permissions, but it does not mean the systems work across retailers or have become a single industry standard.
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
Who Controls the Agent Checkout
The practical issue is not only whether an agent can find a low price. It is who mediates the transaction, what information that party receives, and which company retains the relationship with the shopper. If a customer asks an independent agent to compare products, the agent may direct a purchase without the retailer’s usual advertising or loyalty channels. If the retailer requires shoppers to use its own assistant, it retains more control over the shopping experience and the resulting customer data.
That shift could affect companies that depend on referral traffic as well. The source reports that Amazon Associates commissions rely on customers using tracked links. An agent that reads a review but buys through another route may leave no click for a publisher to receive credit for. HUMAN Security’s September figures, as cited in the source, show 81% of agent requests went to product and search pages and 2.4% to checkout and payment. Those are shares of observed agent requests, not a measure of all online shopping, and the source gives no broader comparison baseline. They suggest that agent activity is currently concentrated more in research than checkout.
For shoppers, the convenience comes with questions about authorization and accountability. Walmart’s spending controls address one part of that concern, but consumers still need to understand what an agent may purchase, how charges are handled and which retailer’s return rules apply. The structure of the transaction may matter as much as the agent’s product recommendation.
Popular right now
Samsung 48-Inch S90H OLED 4K Smart TV (2026)
Bose QuietComfort True Wireless Earbuds with Active Noise Cancellation, Black
Philips 3300 Series Fully Automatic Espresso Machine with LatteGo Milk System
As an affiliate, we earn on qualifying purchases.
A Fast-Moving Retail Divide
The recent sequence of announcements has been compressed. Meta launched Muse on September 8, describing it as an agent that can browse, fill out forms and act for a user. The company said Muse runs in a dedicated virtual machine in its cloud and uses Link by Stripe for payments, including a one-time-use card. Amazon blocked Muse on September 20. Meta announced retail connections with Best Buy, Gap, Sephora, Walmart, Wayfair and Expedia three days later.
On September 28, Shopify said browser agents could complete checkout using structured tools rather than relying only on screen-based interactions. In October, Meta and Sierra announced the Poppy protocol with Walmart, Shopify and Stripe, then published a draft with additional design partners. Walmart’s October 21 announcement and Meta’s adoption figures brought the retailer-versus-agent question into sharper focus. The developments remain separate initiatives: Poppy, Google’s Universal Commerce Protocol and OpenAI and Stripe’s Agentic Commerce Protocol have different backers and scopes, and the source does not establish that they are interchangeable.
There are three broad approaches in play: agents can act through ordinary websites, retailers can provide an approved technical entry point, or companies can adopt shared protocols for identity and permissions. Browser-based shopping can reach many sites, but retailers may object that it is hard to distinguish an agent from a human customer. Structured connections can offer clearer controls, but only where retailers and agent providers support them.
“Muse “can open a browser, fill out forms, and negotiate on their behalf.””
— Meta, in its Muse launch post
Permissions, Payments and Attribution
The scale and reliability of agent shopping remain uncertain. Meta’s figures are company-reported, and the source does not specify how the average purchase value was calculated or independently validate the number of users. The cited request data show browsing activity is more common than checkout, but they do not establish how often agents complete purchases across the wider market.
It is also unclear how widely retailers will accept outside agents, how spending limits will work across different services, and who will be responsible when an agent makes an unwanted purchase. Amazon and Meta give conflicting accounts of the agent’s handling of credentials; the material provided does not independently resolve that dispute. Nor does it establish which commerce protocol, if any, will gain broad adoption or how affiliate commissions will be assigned when an agent influences a purchase without a tracked link.
Retail Access and Standards Ahead
The next test is whether announced integrations and protocols become usable services across a meaningful range of retailers. Walmart’s Agent Gateway and Agent Profiles will need to show how merchants approve agents, set shopping permissions and handle transactions. Protocol developers will need to demonstrate how their systems work together, particularly around identity, consent and payment authorization.
For shoppers, the near-term decision is whether to grant an agent permission to act and what limits to set. Meta, retailers and payment providers may publish further details as their tools develop. Until then, the central questions remain practical: which shops an agent can access, what it can spend, and which company is accountable for the purchase.
Key Questions
What did Walmart announce?
Walmart announced an Agent Gateway for approved agents to search inventory and make purchases through a structured connection, along with Agent Profiles that can define permissions and spending limits.
Why did Amazon block Meta’s Muse?
Amazon said Muse was an unauthorized agent, had not identified itself and appeared to capture and store customer credentials. Meta disputed the credential concern, saying Muse cannot see users’ passwords or payment methods.
How many people have set up Muse?
Meta said on its October 21 third-quarter call that more than 10 million people in the US and Canada set up Muse in about six weeks. The figure is Meta’s own report and is not independently verified in the source material.
Are AI agents already completing many purchases?
The source cites HUMAN Security data from September showing 81% of observed agent requests went to product and search pages, compared with 2.4% to checkout and payment. That indicates more activity in research than checkout in that dataset, not the overall share of online purchases made by agents.
What should shoppers check before allowing an agent to buy?
Shoppers should review the agent’s purchase permissions and spending caps, understand how payment is authorized, and check which retailer’s return policy applies. The source also suggests using approval for each purchase or a one-time-use payment card where available.
Source: ThorstenMeyerAI.com
