🔍 Read the full analysis: AI Shopping Agents: What Amazon’s Block And Walmart’s Welcome Reveal 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 United States and Canada set up its Muse agent in about six weeks, with shopping its most common use. Amazon blocked Muse, while Walmart announced agent profiles and a gateway for approved agents, laying out competing approaches to AI-assisted shopping. The scale of actual agent purchases, the rules for commissions and the safeguards for shoppers remain unsettled.
Amazon has blocked Meta’s Muse from shopping on its site, while Walmart is opening a route for approved AI agents to search its inventory and make purchases. The split, alongside Meta’s report that more than 10 million people in the United States and Canada set up Muse in about six weeks, shows retailers taking different positions as software agents begin handling shopping tasks.
Meta launched Muse on September 8, describing it as a personal agent that can use a browser, fill in forms and act for a user. A reporter for The Verge later asked Muse where granola bars were cheapest per piece. Muse returned Sam’s Club, Walmart and Costco, then said it could not check Amazon because the retailer had blocked it. The example illustrates a basic limit: an agent’s recommendations depend on which stores permit it to access their sites or services.
Amazon’s stated objections concern authorization, identification and customer credentials. After blocking Muse on September 20, the company said Meta had not told it the agent would shop on Amazon and that Muse did not identify itself. Amazon also said the agent “appears to capture and store customer credentials.” Meta’s launch post said Muse “has no visibility into people’s passwords or payment methods.” Those are the companies’ contrasting descriptions; the source material does not establish an independent finding about how Muse handles credentials.
Walmart announced Agent Profiles and a Walmart Agent Gateway on October 21. The gateway is based on the Model Context Protocol, or MCP, and is designed to let approved agents access shopping functions through a structured connection rather than behave like a person clicking through web pages. Agent Profiles are intended to give agents permissions and spending limits. The available material does not specify all participating agents, the full limits on purchases or how the system will handle disputes.
Meta reported its Muse figures during its October 21 earnings call: more than 10 million people in the US and Canada had set up the agent in roughly six weeks, shopping was its most common use, and the average purchase was over $200. These are Meta’s figures, not independently verified sales data. The material does not define whether “average purchase” means completed agent checkouts, purchases associated with Muse more broadly, or another measure.
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 open a product page. Retailers and agent developers are negotiating who controls the shopping relationship, payment path and purchase data. If a customer asks an assistant to compare products and then buy, the agent may shape where the customer goes—and which retailer receives the sale.
That creates a potential challenge for advertising and referral businesses. Amazon reported more than $68 billion in advertising revenue last year, according to GeekWire, while Meta’s Mark Zuckerberg said the company expected to earn “a small fee from transactions.” Those figures point to different possible ways companies could benefit from agent-led shopping; they do not establish how much revenue agents will generate or which business model will prevail.
For publishers and affiliate sites, a purchase made without a tracked link could complicate commission attribution. Amazon Associates commissions are tied to customers clicking tracked links, according to the source material. An agent that reads a review but sends a shopper elsewhere may leave no referral click to credit. The size of that effect is not yet known.
There is also a practical difference between product research and checkout. HUMAN Security’s September data found that 81% of agent requests went to product and search pages, while 2.4% went to checkout and payment. The figures describe requests in that dataset and time period; they are not a universal measure of all shopping agents. They suggest that browsing activity should not be mistaken for a large volume of completed purchases.
Products Worth Considering
Sonos Era 100 Wireless Smart Speaker, Black
Samsung 32-Inch Odyssey OLED G8 (G80SH) 4K Gaming Monitor
LG UltraGear 27-Inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDMI 2.1
As an affiliate, we earn on qualifying purchases.
A Short Race to Build Agent Access
The sequence of announcements has moved quickly. Meta introduced Muse on September 8 and announced retail connectors with Best Buy, Gap, Sephora, Walmart, Wayfair and Expedia at its September 23 Connect event. Shopify announced on September 28 that browser agents could use structured tools to complete checkout, rather than relying only on screen interaction.
On October 6 and 9, Sierra and Meta announced the Personal Agent Protocol, also called Poppy, with Walmart, Shopify and Stripe. Its draft is intended to help identify an agent, record what a customer has allowed and carry a session across websites, APIs and company agents. The source describes the draft as version 0.1 and says payment details were left to extensions; the protocol’s eventual scope and adoption remain open.
Other standards are also in play. Google and Shopify are associated with the Universal Commerce Protocol, while OpenAI and Stripe developed the Agentic Commerce Protocol for checkout between an agent and a merchant. Protocol publication does not guarantee widespread use. OpenAI stepped back from its earlier Instant Checkout approach in March 2026, saying its initial version lacked the flexibility it wanted and that it would let merchants use their own checkout experiences while focusing on product discovery.
Agents can also shop through ordinary browser interfaces, through a retailer’s own agent connection or through shared protocols. Browser-based shopping can work on sites built for people, but it may be harder for retailers to distinguish an agent from a customer. Structured connections can give retailers more control over permitted actions, but require merchants and agent providers to agree on access and technical standards.
““Continued access by an unauthorized AI agent violates Amazon’s Conditions of Use, to which our customers have agreed.””
— Amazon
Sales, Permissions and Liability Remain Open
Meta’s audience and purchase figures are company-reported, and the available information does not define the purchase metric or provide an independent breakdown of completed Muse checkouts. It is also unclear how many retailers will support Walmart’s gateway, how broadly Meta’s announced connectors will work, and whether shoppers will be able to use agents consistently across competing services.
The safeguards and responsibilities are unsettled as well. Walmart has described profiles and spending limits, but the material does not provide their full operating rules. It does not explain who would resolve a disputed order, handle a return or bear responsibility if an agent exceeds a user’s instructions. Amazon and Meta also disagree in their public descriptions of Muse’s access to credentials; no independent technical finding resolving that dispute is provided here.
Finally, published protocols do not settle which company will control discovery, checkout or payment records. The source material describes several competing standards and partnerships, but gives no evidence that one has become a shared industry default.
Watch Retailer Rollouts and Agent Checkout
The next test is whether Walmart’s announced gateway and profiles become usable at meaningful scale, and which agents receive access. Retailer participation, the permissions offered to customers and the handling of payment and returns will show whether structured access can turn shopping assistance into dependable transactions.
Meta’s retail connectors and the competing commerce protocols will also need to show how they work beyond announcements and drafts. Further company updates may clarify the number of completed purchases, fees and referral arrangements. Until those details are available, the clearest confirmed picture is a divided market: Amazon has blocked Muse, Walmart has announced an agent-access route, and the rules for agent-led buying are still being negotiated.
Key Questions
Why did Amazon block Meta’s Muse?
Amazon said Muse was an unauthorized AI agent, did not identify itself and appeared to capture and store customer credentials. Meta said Muse had no visibility into users’ passwords or payment methods. The supplied material does not include an independent finding resolving the disagreement.
What is Walmart’s Agent Gateway?
Walmart announced a gateway based on the Model Context Protocol (MCP) for approved agents to access shopping functions through a structured connection. Walmart also announced Agent Profiles intended to set permissions and spending limits. The full operating rules and rollout details are not provided.
How many people use Meta Muse?
Meta said more than 10 million people in the United States and Canada set up Muse in about six weeks. The figure was reported by Meta during its October 21, 2026, earnings call and is not identified in the source as an independently verified count of active users.
Are AI agents already completing most online purchases?
The available evidence does not show that. HUMAN Security’s September dataset found 81% of agent requests went to product and search pages and 2.4% to checkout and payment. Those figures apply to that dataset and period, not necessarily to all agents or shopping activity.
What should shoppers check before letting an agent buy?
Users should check the agent’s purchase approval settings, spending cap and payment method, and confirm which retailer’s return policy applies. The source material advises considering approval for each purchase, a capped budget and a one-time-use card; it does not describe one standard protection across agents.
Source: ThorstenMeyerAI.com
