The landscape, with caveats

Industry coverage describes several overlapping efforts. Google announced its Universal Commerce Protocol in January 2026, with retailers and platforms such as Walmart, Target, Shopify, Etsy, and Wayfair reported among partners, and it works with Google's Agent Payments Protocol for scoped payment authorization. OpenAI and Stripe released the Agentic Commerce Protocol as open source in September 2025, and reports say OpenAI removed in-chat checkout in March 2026, citing data quality problems with early merchants, though sources differ on what that means for the protocol itself. The Model Context Protocol often supplies the data connection to inventory and pricing.

Most of this coverage comes from vendors and agencies, and adoption and sales-uplift claims trace back to a few repeated figures. Treat them as unverified.

The early signal from the checkout reversal is instructive: agents need accurate, current product and availability data, and merchants' data was often not good enough.

What to invest in

Build a clean internal product and inventory API as the backbone: accurate titles, attributes, prices, availability, shipping options, and return policies, updated in near real time. Agents punish stale or ambiguous data far more harshly than human shoppers do.

Then add thin adapters for each protocol you decide to support. If one gains traction, you add it; if one fades, you retire it. That avoids rebuilding your commerce stack around a bet.

Fix the data first

Complete, accurate, structured product data is the common requirement of every agent protocol.

Adapters over rewrites

Keep protocol-specific code at the edge so the core commerce system does not change.

Plan for fraud, returns, and attribution

Agent-initiated purchases raise questions your current systems may not answer: how to verify the agent acts for a genuine customer, how to apply limits, how to handle disputes and returns, and how to attribute sales to channels. Decide your policies and instrument your analytics before volume arrives.

Start with a small pilot on one protocol and a limited catalog to learn the operational reality.

Preparing a store for AI shopping agents? Talk to us about a product-data and integration plan.

Key takeaways

  • Several protocols compete, including Google's UCP and AP2 and OpenAI and Stripe's ACP; reports on specifics differ.
  • Reported early merchant data quality problems show agents demand accurate, current product data.
  • Build a clean internal product and inventory API with thin adapters per protocol.
  • Prepare fraud, returns, and attribution policies and run a small pilot.
This is general information, not business or legal advice. It draws on vendor and industry coverage as of October 2026; verify protocol specifications and status with the primary documentation.