Walmart Agent Gateway: AI Shops for You Automatically—Why Is Amazon Going the Opposite Way?

Walmart recently launched Agent Gateway, allowing third-party AI assistants like ChatGPT, Gemini, and Meta AI to connect directly to its product catalog and handle search, price comparison, and even add-to-cart on users' behalf. Meanwhile, Amazon is blocking external AI agents. The two retail giants are heading in completely opposite directions—a battle over traffic gateways and a redistribution of control in the AI e-commerce era.

What Can Agent Gateway Do?

According to a demo by the YouTube channel "AI Dost India," a user simply tells the AI assistant "find me earphones under 2,000," and the agent automatically searches Walmart products, compares specs and prices, and adds candidates to the shopping cart. The entire process requires no opening of the Walmart app or website—the chat window becomes the new shopping gateway.

The key to this flow is "acting on behalf" rather than "deciding on behalf." The AI handles information retrieval and filtering, but the user still confirms the final purchase. The channel repeatedly emphasized one design principle in the video: control stays in the user's hands.

AI Agent refers to an intelligent program capable of autonomously perceiving its environment, making decisions, and executing actions. Unlike traditional chatbots, agents can not only answer questions but also call external tools and APIs to complete multi-step tasks—such as searching for products, comparing prices, and adding items to a cart. In e-commerce scenarios, agents access product data through interfaces provided by Walmart, simulating user actions, but the final transaction confirmation remains with humans. This "human-AI collaboration" model leverages AI's efficiency advantages while preserving user control—a typical form of current agent technology implementation.

Three Safety Valves: Limits, Approvals, Independent Tracking

Letting AI handle money directly—security and trust are the biggest hurdles. Walmart's Agent Gateway provides three layers of protection:

Self-set limits. Users can set per-transaction or cumulative spending caps for the AI agent; operations exceeding the limit cannot execute. This is essentially installing a "wallet gate" on the agent.

Per-transaction approval. Every purchase requires manual user confirmation before finalization—the agent cannot bypass this step to pay directly.

Independent tracking. All operations and transaction records generated by the AI agent are tracked separately from the user's own shopping behavior, facilitating post-hoc auditing and error correction.

These three design choices address users' concerns about "AI spending recklessly." From a product logic perspective, Walmart positions the agent as an "assistant with permission boundaries" rather than a "fully authorized proxy." This conservative stance is reasonable in the early stages—trust must be built gradually.

In the architecture of AI agents connecting to e-commerce platforms, API Gateway plays a key role. Agent Gateway is essentially a controlled interface layer that allows third-party AI assistants to access Walmart's product catalog and shopping cart functions in a standardized way, while implementing permission management, traffic control, and audit logs. Unlike directly opening up databases, a gateway can finely restrict what agents can and cannot do—for example, prohibiting direct payments or limiting query frequency. This design both ensures platform data security and provides external developers with a usable access point—a balance between open ecosystem and risk control.

Why Is Amazon Going the Other Way?

In stark contrast to Walmart's open posture, Amazon is blocking external AI agents from scraping and operating its product data. According to analysis from the same source, Amazon's strategy is to keep AI shopping capabilities within its own system rather than letting third parties like ChatGPT become new traffic distributors.

Both approaches have their logic. Walmart, as the challenger, is willing to open interfaces in exchange for traffic and exposure brought by AI platforms; Amazon, as the incumbent, cares more about user relationships and exclusivity of transaction data. If users could complete Amazon shopping within ChatGPT, Amazon would degenerate into a backend fulfillment system, losing direct touchpoints with consumers.

The essence of this divergence is: will AI assistants become the new gateway to e-commerce? Walmart bets "yes" and stakes its claim early; Amazon bets "won't give it up easily" and tightens the gates.

Traffic gateway refers to the starting point of a user's shopping journey. In the traditional e-commerce era, the gateway was the platform app or website; in the AI era, conversational assistants may become the new gateway. If users become accustomed to describing needs in ChatGPT and completing purchases directly, e-commerce platforms will recede into backend suppliers, losing opportunities for direct interaction with consumers—and thus struggling with advertising recommendations, membership operations, and brand building. Therefore, controlling the gateway means commanding user attention and transaction data. Walmart's opening of Agent Gateway aims to leverage AI platforms for new traffic, while Amazon's blocking of external agents is about defending its own gateway and avoiding being piped.

India Market: Time Gap and Preparation

Agent Gateway is currently only available in the US. The channel host judges that similar features will come to India sooner or later, at which point Flipkart and Amazon India may both follow suit.

For Indian users, this window period is a good time to build habits. The video's advice is straightforward: regardless of which platform the shopping agent comes from, spending limits and per-transaction approval must always stay in your own hands. This isn't a technical issue—it's a habit issue. When agents become smarter and operations smoother, the two defenses users are most likely to relax are precisely these.

Three Changes Once Shopping Agents Become Mainstream

First, search behavior will shift from "browsing shelves" to "describing needs." Users will no longer flip through pages comparing prices but instead give constraints in natural language, with agents executing. This places higher demands on product titles and the structured-ness of specifications.

Second, platform competition will extend from "product richness" to "agent friendliness." Whoever has more open interfaces, more standardized data, and smoother approval flows will be more likely to be selected by AI assistants.

Third, user security awareness will become part of product design. Limits and approvals are not optional—they are prerequisites for shopping agents to be trusted. The fact that Walmart presents these three safety valves as the first layer of its feature announcement speaks volumes.

The fork between Walmart and Amazon has only just begun. Whether shopping agents become the mainstream gateway depends on whether users are willing to hand "browsing" over to AI—and whether platforms are willing to return control to users.

One-line Summary

Walmart opens AI agent access to its product catalog while Amazon blocks it—the rivalry reveals who controls the new e-commerce gateway.

Paragraph Summary

Walmart launched Agent Gateway, allowing third-party AI assistants like ChatGPT and Gemini to connect directly to its product catalog for search, price comparison, and add-to-cart. The feature has three safety valves: user-set spending limits, manual approval for each purchase, and independent tracking of agent operations. In contrast, Amazon blocks external AI agents, keeping AI shopping capabilities within its own system. The divergence between the two giants is essentially a battle over whether AI assistants can become the new e-commerce gateway. Walmart, as challenger, trades openness for traffic; Amazon, as incumbent, tightens the gates to protect user relationships and data exclusivity. The feature is currently US-only, with India expected to follow, and users should always retain control over limits and approvals.