# Walmart launches Agent Gateway MCP server and agent spending controls

published: 2026-10-22T08:05:40.256Z
topic: mcp
source: https://techcrunch.com/2026/10/21/while-amazon-blocks-ai-agents-walmart-rolls-out-the-welcome-mat/
url: https://aiagentnews.io/stories/walmart-launches-agent-gateway-mcp-server-and-agent-spending-controls

Walmart has launched Agent Gateway, an MCP server that lets approved AI agents search its online inventory and make purchases for shoppers. The retailer is also adding Agent Profiles with per-transaction approval or spending limits, along with new terms that govern how agents may buy on its site.

Walmart announced the launch on Wednesday, positioning it against Amazon, which has begun blocking AI agents from buying on behalf of online shoppers. Walmart already had a partnership with Meta's Muse and integrations with ChatGPT and Gemini. The Agent Gateway uses the Model Context Protocol, originally created by Anthropic, to connect approved agents to Walmart.

Each agent can be authorized with its own Agent Profile in a customer's account, so its purchases appear separately from the customer's own. Customers can require approval for every transaction or allow purchases within a spending limit they set. Profiles roll out next month and are created the first time an agent shops through the gateway, and agent support is open to any Walmart account holder without a Walmart+ membership.

Walmart's new terms say its returns process applies even when an agent does not act as intended. Agents cannot act beyond what the shopper intends, cannot access payment information, and can only see the data needed for a task. Agents that do not follow the terms will be blocked from the site.

## Why it matters if you build agents

Agents can now search Walmart's inventory and buy through a standard MCP connection, but only when the agent is approved and operating under Walmart's terms. Builders should expect the retailer to enforce per-agent profiles and spending caps, so agents need to handle approval steps and limits gracefully. Agents that break the terms will be cut off, so keep purchase actions tightly scoped to what the user asked for.