Realtime AI News
As Amazon blocks AI agents, Walmart opens the door with an Agent Gateway
Walmart launched an Agent Gateway MCP server that lets approved AI agents search its inventory and buy on a shopper's behalf, alongside Agent Profiles that give each agent its own identity, spending limits and per-transaction approval options. The move contrasts with Amazon, which recently began blocking agents from making purchases.

Walmart announced a set of technologies for agentic commerce on Wednesday, anchored by a new MCP server called Agent Gateway. President and CEO John Furner unveiled the plans at the Masters of Scale Summit in San Francisco, saying the technology will support agentic shopping for both Walmart and Sam's Club.
Agent Gateway is built on the Model Context Protocol, the standard originally created by Anthropic for connecting agents to external data sources, tools and workflows. In this case it lets an approved agent search Walmart's online inventory and complete purchases for the person it represents.
Walmart says the gateway can help customers buy their usual brands or items without handing their personal information or purchase history directly to the agent. That is an attempt to answer the most practical objection to agentic shopping: a useful agent needs a lot of personal context, and shoppers rarely want to hand all of it over.
The retailer also built technology to tell an agent's purchases apart from a customer's own. Agent Profiles, rolling out next month, create a dedicated profile in a shopper's account for each agent and show its transactions separately; shoppers can require approval for every transaction or set a spending limit that authorizes purchases automatically.
Walmart compares the controls to adding an authorized user to a credit card: transactions are identified and spending can be capped. A profile is created the first time a shopper uses a given agent through the gateway, can be deleted at any time, and requires only a Walmart account, not a Walmart+ membership.
Alongside the tooling, Walmart published terms for AI agents that specify how they must operate and what rights they have to transact. Returns apply even when an agent does not act as intended, agents cannot go beyond what the shopper intends or access payment information, and agents that do not comply get blocked from the site.
The contrast with Amazon is the story. Amazon recently began blocking agents from making purchases on shoppers' behalf, and travel and booking sites have locked agents out entirely; Walmart had already partnered with Meta's Muse and integrated with ChatGPT and Gemini, and is now building the underlying rails.
Opening the door is not the same as letting go. Walmart is trading a traceable account structure, explicit terms and hard limits for agent-driven sales, which is roughly the bargain retail will have to strike if agentic checkout is going to become ordinary. The near-term test is how Agent Profiles actually roll out next month, and whether agents operating under those terms convert.
Why it matters
Agentic shopping now has something close to a real payment framework: account separation, per-transaction approval, spending caps and a returns backstop. The retailer with the clearest rules is the one most likely to capture agent-driven orders.
Nearby Updates
All10/22, 03:50
GPT-5.6 Sol and Gemini 4 Argon found a critical PHP flaw affecting millions of sites
OpenAI's GPT-5.6 Sol discovered a critical PHP vulnerability through the Patch-the-Planet initiative, and security firm Wiz independently found the same flaws using Wiz Atlas with Gemini 4 Argon and GPT-5.6 Sol. The fix took two months of coordinated work across five phases, and OpenAI warns the same capabilities can help attackers.
10/22, 03:53
Anthropic launches Claude Duet, an always-on agent that comes to you first
Anthropic opened a research preview of Claude Duet, an app built around Claude Opus 5.5 in which Claude acts as a proactive agent that monitors connected apps, works on a cloud virtual machine and checks in before sending messages, making payments or deleting anything. It is included in Pro and Max plans for US subscribers via a waitlist, and is not yet available on Team or Enterprise plans.
10/22, 04:27
Recursive ships Flint, a free coding model built largely by its own AI
Recursive released Flint, a free coding model that runs on a laptop and, according to the company, was built largely by its automated research system, the Eureka Machine. Flint is a 40-billion-parameter mixture-of-experts model with 2.5 billion active parameters, published under Apache 2.0 on Hugging Face.
10/22, 02:26
NVIDIA releases Kumo Relational, a no-training foundation model for relational data
NVIDIA published Kumo Relational on the Hugging Face blog, a pre-trained foundation model that predicts directly on enterprise relational databases without task-specific training or feature engineering. Across 41 tasks on 15 databases it beats supervised relational learning and expert feature pipelines, and on the SAP SALT enterprise benchmark it tops AutoGluon ensembles by 8% in mean reciprocal rank.