🔍 Read the full analysis: A Closer Look At Amazon And Walmart’s Competing AI Shopping Policies 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
Walmart has announced an Agent Gateway and spending controls for AI shopping agents, while Amazon has blocked Meta’s Muse agent and requires agents to identify themselves. The competing approaches reflect a broader fight over who controls shoppers, checkout and purchase data; the practical reach of agent shopping remains limited and uneven.
Walmart has announced an Agent Gateway for approved AI shopping agents and customer profiles with spending limits, while Amazon has blocked Meta’s Muse and says agents must identify themselves under its Conditions of Use. The opposing policies put two major retailers on different sides of a growing contest over who controls the shopping relationship when an AI agent researches or buys a product for a person.
Walmart announced its Agent Gateway and Agent Profiles on October 21. The gateway is based on the Model Context Protocol, or MCP, and is intended to let approved agents search Walmart inventory and make purchases through a structured connection rather than by clicking through consumer webpages. Agent Profiles are described as a way to set permissions and spending limits. The source material does not specify which outside agents can use the gateway now or how its controls will work in practice.
Amazon, by contrast, cut off access by Meta’s Muse on September 20. Amazon said Muse had not identified itself as an agent, that Meta had not told the retailer it would shop on Amazon, and that the agent appeared to capture and store customer credentials. Meta disputed the credential concern: its launch post said Muse “has no visibility into people’s passwords or payment methods.” Amazon’s customer-facing notice said continued access by an unauthorized AI agent violated the site’s Conditions of Use.
Meta reported on an October 21 earnings call that more than 10 million people in the United States and Canada had set up Muse in about six weeks, and said shopping was its most common use. Meta also reported an average purchase above $200. Those are company figures, not an independent measurement of completed agent purchases. Separate data from HUMAN Security, cited in the source material, found that in September 81% of agent requests went to product and search pages, while 2.4% went to checkout and payment pages. The figures suggest that browsing is currently more common than agent-led checkout, though they do not establish how often shoppers ultimately buy.
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 policy split matters because an AI agent can influence which products a shopper sees, where they buy and which retailer gets the sale. If a retailer opens a structured connection, it may make its inventory and checkout easier for an agent to use. If it blocks agents or requires identification, it can limit access and set conditions for how automated shopping happens. Walmart’s approach creates an authorized route; Amazon’s stance keeps tighter control over access to its store and promotes its own assistant.
There are also consequences for advertising and referral businesses. The source material says Amazon earned more than $68 billion from advertising last year, citing GeekWire, and reports that Meta wants to earn “a percent of the value” Muse creates. Those interests make control of the agent commercially important, but the precise economics of the new tools are not disclosed. Affiliate publishers may also lose credit if an agent reads a review but completes a purchase without using a tracked link. Amazon Associates terms tie commissions to tracked clicks, according to the source. How sales are attributed could affect retailers, platforms and publishers alike.
For shoppers, convenience comes with practical questions about permission, payment and responsibility. A spending cap or approval requirement can limit unintended purchases, but the source does not establish what protections Walmart’s profiles provide or how disputes will be handled across different agents. Consumers may also need to know which company’s returns and customer-service policies apply when an agent places an order.
Products Worth Considering
Sonos Era 100 Wireless Smart Speaker, Black
DJI Osmo Mobile 8 Standard Combo Phone Gimbal Stabilizer
Amazon Kindle Paperwhite Signature Edition (12th Generation, 2024), 32 GB, Metallic Black
As an affiliate, we earn on qualifying purchases.
From Browser Bots to Retail Gateways
The current disagreement follows a rapid series of announcements. Meta introduced Muse on September 8, describing it as a personal agent able to open a browser, fill out forms and act for users. The source says Muse operates in a dedicated virtual machine in Meta’s cloud and uses Link by Stripe, which issues a one-time-use card. After Amazon blocked Muse, Meta announced retail connectors with Walmart and other brands on September 23. Walmart’s October 21 gateway announcement came as Meta shared its usage figures.
Retailers and technology companies are also developing different ways for agents to interact with stores. Shopify has introduced structured checkout tools, while Meta, Sierra, Walmart, Shopify and Stripe are among the backers of the draft Personal Agent Protocol, also called Poppy. Google and Shopify are associated with the published Universal Commerce Protocol, and OpenAI and Stripe have an Agentic Commerce Protocol. These efforts are not one settled standard: they cover different parts of shopping and have different participants. The source notes that OpenAI shifted its Instant Checkout approach in March 2026, saying merchants could use their own checkout experiences while the company focused on product discovery.
Agents can also shop through ordinary webpages, acting much like a person using a browser. That route can work without a retailer-built connection, but it can make automated traffic harder to distinguish from human activity. Structured gateways and protocols offer another route, with retailers defining how agents identify themselves and interact with inventory or checkout. The dispute is consequently about both technical access and commercial control, not just whether an agent can navigate a website.
“Continued access by an unauthorized AI agent violates Amazon’s Conditions of Use, to which our customers have agreed.”
— Amazon, in a notice shown to Muse users
Access, Safeguards and Sales Data
Several practical details remain unsettled. The available material does not say which agents Walmart has approved for its gateway, what data those agents can access, or whether customers must approve each purchase. It also does not provide independent evidence for Meta’s reported user and purchase figures, or define how the reported average purchase was calculated. The scale of completed agent purchases is therefore unclear.
The accounts of Amazon’s dispute with Meta are also competing statements: Amazon raised concerns about identification and credentials, while Meta said Muse could not see passwords or payment methods. The source does not provide technical findings that resolve that disagreement. It is also unclear how retailers will allocate responsibility for mistaken orders, returns, refunds or customer support when an outside agent is involved. Protocol adoption and referral attribution remain open questions as well; announced standards and partnerships do not establish that merchants will adopt them widely.
Retailer Adoption and Agent Rules
The next test is whether Walmart’s gateway moves from an announcement to broad use by outside agents, and what permissions and purchase safeguards accompany it. Amazon’s policy may also develop as it clarifies how agents must identify themselves and as its own shopping assistant competes for use. The companies’ published rules and any changes to agent access will show whether the two approaches remain distinct or converge around shared requirements.
For shoppers and businesses, useful signals will include clearer disclosures about purchase approval, spending caps, payment handling, returns and the recording of referrals. The draft Poppy protocol and other commerce standards may be revised, but their current status does not establish which will become widely used. For now, agent shopping is developing unevenly: some systems are building dedicated connections, while others restrict automated access, and the extent of routine agent-led checkout has not been established.
Key Questions
What did Walmart announce?
Walmart announced an MCP-based Agent Gateway for approved shopping agents and Agent Profiles intended to let customers set permissions and spending limits. The available information does not detail current agent availability or all profile settings.
Why did Amazon block Meta’s Muse?
Amazon said Muse had not identified itself as an agent and appeared to capture and store customer credentials. Meta said Muse could not see users’ passwords or payment methods. The source material does not include independent technical findings resolving the disagreement.
Are AI agents already completing many purchases?
The cited September HUMAN Security data says 81% of agent requests went to product and search pages and 2.4% went to checkout and payment pages. That measures requests, not the share of purchases made by agents, so the overall volume of completed agent purchases remains unclear.
What should shoppers check before using an agent?
Shoppers should check what the agent is allowed to do, whether purchases require approval, and whether spending limits or a one-time-use payment card are available. They should also confirm which retailer’s return and customer-service rules apply to an order.
Source: ThorstenMeyerAI.com
