Walmart Opened Its Agent Gateway — And 25 Billion AI Requests Per Quarter Show Why Every Retailer Will Follow

Walmart Opened Its Agent Gateway — And 25 Billion AI Requests Per Quarter Show Why Every Retailer Will Follow

Jasper Koers 11 min read Brand Intelligence

On October 21, 2026, Walmart did something no major retailer had done before: it opened an official gateway for AI agents to shop on behalf of consumers. The Walmart Agent Gateway is an MCP server — a machine-readable interface built on the Model Context Protocol — that lets approved AI agents search Walmart's inventory, add products to a cart, make purchases, and track orders. Meta's Muse, OpenAI's ChatGPT, and Google's Gemini are named as launch partners.

Three days earlier, DataDome published its Q3 2026 AI Traffic Report. AI agent traffic on its network hit 24.9 billion requests for the quarter — up 41.5% from 17.7 billion in Q2. Agentic browser traffic — agents that navigate websites like a human with a browser — jumped 4.7x in September alone, from 12.5 million to 58.2 million requests. Even excluding Meta's Muse, which launched in late September and generated 29.1 million requests in its first 12 days, agentic browser traffic still tripled.

These two announcements are not separate stories. They are cause and effect. The traffic surge explains why Walmart built the gateway. The gateway reveals how the rest of retail will respond.

Why Walmart Built a Gateway Instead of Just a Website

Walmart's announcement includes a telling admission: AI agents browsing walmart.com through a browser are "slow, error-prone, and prone to getting stuck." This is the same website that serves 120 million weekly customers without issue. For humans, it works. For agents, it does not.

The problem is fundamental, not fixable with better scraping. E-commerce websites are designed for human cognition — visual hierarchies, promotional carousels, nested filters, dynamic pricing displays. An AI agent navigating this interface is doing the computational equivalent of reading a novel to find a phone number. It works, technically. It is also absurd when you could just look up the number.

The Agent Gateway replaces this with structured, machine-readable access. An agent can query product availability, compare prices, check specifications, and complete a purchase through direct API calls instead of screen-scraping a website designed for a different species of intelligence.

This is the same shift that happened with mobile. Retailers first built mobile-optimized websites, then realized mobile apps with native APIs were fundamentally better. The agent equivalent is happening in compressed time. Walmart skipped the "optimize the website for agents" phase entirely and went straight to the native interface.

How the Gateway Works

The Agent Gateway is built on the Model Context Protocol — the same open standard that Anthropic introduced for AI tool integration. This is significant. MCP is becoming the de facto interface layer between AI agents and business systems, and Walmart's adoption validates the protocol for high-stakes commerce.

Agent Profiles split the agent's identity from the customer's identity. The first time a customer uses an AI agent to shop through the gateway, Walmart creates a profile for that agent within the customer's account. Walmart compares this to adding an authorized user to a credit card. Agent purchases appear separately from the customer's own.

Spending controls give customers three options: require approval for every purchase, set a spending limit ($50, $100, or $200 were given as examples), or allow unrestricted purchasing within the customer's existing payment methods. Customers can revoke an agent's access at any time.

Data isolation is enforced by design. Agents cannot see payment card details. They cannot add or change payment methods. For routine purchases — a customer's usual brand of coffee, their regular paper towels — the gateway finds the right products without sharing the customer's broader purchase history or personal information with the agent.

Enforcement is real. Agents that take actions beyond what the shopper intended get blocked from Walmart's site entirely. This is not a terms-of-service warning. It is an automated enforcement mechanism.

The Amazon Contrast Is Now a Strategic Divergence

Two weeks ago, we covered Amazon blocking Meta's Muse from shopping on its platform. That was a defensive move — Amazon protecting its walled garden from AI agents that would disintermediate its recommendation engine.

Walmart's announcement turns that defensive move into a strategic liability for Amazon. Here is why.

DataDome's Q3 report shows that Meta's crawlers generated 12.4 billion requests in Q3 — 49.7% of all AI agent traffic on DataDome's network. Meta's Muse specifically generated 29.1 million agentic browser requests in its first 12 days of full availability. Of those requests, 99.3% came from cloud infrastructure, not consumer devices. And 8.9% of Muse's requests touched form, login, payment, account-creation, or cart endpoints — the transactional core of e-commerce.

Amazon blocked this traffic. Walmart is building an official interface for it. The brands that sell on both platforms now face a structural divergence: their products are agent-discoverable on Walmart and agent-invisible on Amazon. For categories where consumers are increasingly delegating purchases to AI agents, this divergence will show up in market share data within quarters, not years.

The 25 Billion Request Reality

The DataDome numbers deserve scrutiny because they reveal where agent commerce is heading before it arrives.

25 billion requests per quarter means AI agents are collectively hitting websites roughly 275 million times per day across DataDome's network alone. DataDome protects a fraction of the internet's total traffic, so the real number is multiples higher. This is not experimental traffic. This is infrastructure-scale demand for machine-readable product data.

The 4.7x agentic browser spike in September matters more than the absolute number. Agentic browsers are agents that navigate websites the way a human would — logging in, filling forms, clicking buttons, completing purchases. Their traffic jumping from 12.5 million to 58.2 million requests in a single month signals that the consumer-facing agent commerce wave has started.

ClaudeBot surpassed GPTBot in Q3, ranking fourth among AI agents by traffic volume. The AI agent landscape is diversifying. Brands that optimize for one agent platform are building on an unstable foundation.

Meta drives half the traffic with 12.4 billion requests, but its share dropped from 51.5% to 49.7% because non-Meta agents are growing faster. The ecosystem is expanding. This is not a one-company phenomenon.

What Walmart's Gateway Means for Brand Data

The Agent Gateway does not change what brands need to do. It makes what they should have been doing urgent.

Product Data Is Now an API Contract

When a human shops on Walmart, they tolerate ambiguity. A product title like "Premium Organic Fair-Trade Single-Origin Ethiopian Yirgacheffe Coffee Beans" communicates quality signals to a human even though it is keyword-stuffed nonsense from a data-structure perspective.

When an AI agent shops through the Agent Gateway, it queries structured attributes. It asks for coffee beans, organic, Ethiopian origin, 12-ounce bag. If your product data does not decompose into clean attributes — category, origin, size, certification, brand — the agent either cannot find your product or cannot evaluate it against alternatives.

This is not a future problem. The gateway is in beta now. The agents are already querying.

Brand Identity Must Be Machine-Verifiable

The Agent Gateway creates agent profiles and enforces data isolation precisely because agent commerce requires trust verification at every layer. The agent needs to trust Walmart. The customer needs to trust the agent. And the agent needs to trust the brands whose products it recommends.

Brand trust in an agent-mediated world is not about brand recognition. It is about brand data consistency. When an agent evaluates your product, it cross-references your product description against reviews, your pricing against competitors, your brand information against third-party directories, your specifications against independent testing data. Inconsistencies reduce trust scores. Trust scores determine recommendation probability.

The Three Surfaces That Now Matter

Walmart's gateway launch clarifies which surfaces determine whether an agent finds, evaluates, and recommends your brand:

1. Your product data feed. The structured attributes, specifications, and categorization data that feeds into retail platforms. This is the data the Agent Gateway queries. If it is incomplete, inconsistent, or keyword-stuffed instead of attribute-structured, agents will skip your products in favor of competitors with cleaner data.

2. Your brand data across verification sources. Directories, review platforms, industry databases, business registries. Agents cross-reference these sources to verify brand legitimacy and product claims. A brand with consistent information across 30 verification sources is more trustable to an agent than a brand with a beautiful website and zero third-party presence.

3. Your website's machine-readable layer. JSON-LD schema, structured product data, clear categorization, accurate metadata. Even with the Agent Gateway, agents still crawl websites for brand-level information. The 25 billion quarterly requests confirm that agents are reading your website right now — the question is what they find.

The Retailer Domino Effect

Walmart is the largest retailer in the world by revenue. When Walmart adopts a protocol, the protocol becomes a standard. Every other major retailer now faces a strategic choice: adopt a similar gateway and accept agent-mediated commerce, or follow Amazon's blocking strategy and risk losing agent-driven market share to competitors who do.

Target, Costco, Kroger, Home Depot — they are all watching the Walmart-Amazon divergence play out in real time. The retailers who open their own gateways will capture a disproportionate share of agent-driven commerce in their categories. The retailers who block agents will protect their existing recommendation engines at the cost of a growing channel.

The smart bet is that most will follow Walmart. MCP is an open standard. Building an Agent Gateway is not a multiyear infrastructure project. It is months of engineering work, and the protocol handles the hard parts — authentication, permission management, data isolation.

When they do, brands will face an agent-accessibility gap. Products with clean, structured, attribute-level data will be queryable through every gateway. Products described in marketing prose with keyword-stuffed titles will not translate to the structured queries that gateways process.

What Brands Should Do Now

Audit your product data for attribute completeness. Pull your product listings from every retail platform you sell on. For each product, check whether every relevant attribute is populated — not just title and price, but category, specifications, materials, dimensions, certifications, target audience. Missing attributes mean missing from agent query results.

Separate structured data from marketing copy. Your product descriptions should serve two purposes independently: human-readable marketing copy and machine-readable attribute data. If the only place your product's origin, material, or certification appears is buried in a paragraph of marketing prose, agents cannot query it.

Ensure brand data consistency across platforms. An agent cross-referencing your brand will check your website, your retail platform listings, your directory profiles, your review platform presence, and your social media descriptions. Inconsistencies in brand name, product categories, pricing, or business information degrade trust scores.

Monitor your AI agent traffic now. Check your server logs for AI crawler user agents — ClaudeBot, GPTBot, Meta-ExternalAgent, Amazonbot, and others. The 25 billion quarterly requests mean agents are already reading your product pages. What they are finding today determines whether they recommend you tomorrow.

Prepare for multi-gateway commerce. Walmart's gateway is first, not last. Structure your product data in a format that translates cleanly to any MCP-based interface. Clean, attribute-level product data is gateway-agnostic. Marketing-optimized descriptions are gateway-hostile.

The Gateway Is Open. The Question Is Whether Your Data Walks Through It.

Walmart opened the first major retailer gateway for AI shopping agents. The 25 billion AI requests per quarter confirm that agent traffic is already at infrastructure scale. Meta's Muse hit 29 million agentic browser requests in 12 days. ClaudeBot surpassed GPTBot. The ecosystem is growing, diversifying, and transacting.

The brands that win agent-mediated commerce will not be the ones with the best marketing copy or the most creative product photography. They will be the ones whose product data is clean, structured, consistent, and queryable through any interface an agent uses — whether that is Walmart's gateway, a future Target gateway, or an agent navigating your website directly.

The gateway is open. Your product data is the ticket. Make sure it is machine-readable before the agents come shopping.

Make your brand agent-ready. Extract your brand fingerprint and verify your product data is structured, consistent, and queryable across every platform where AI agents shop.

Share this article

Ready to try the API?

Extract brand data from any website with a single API call. Start free.