Agentic commerce, for store owners.
Agentic commerce means an AI assistant doing part of the shopping: finding a product, comparing options, and in some cases completing the purchase. Two open protocols now define how stores connect to it. Most of what a store needs to do today is less exotic than the name suggests.
- Published
- 22 October 2026
- From
- Kramiva
01
What it is
A shopper asks an assistant such as ChatGPT or Google's Gemini for a product. Instead of returning ten links, the assistant reads product data from stores, compares it, and presents a few options. Depending on the platform and the store, the purchase can then happen on the store's own checkout or inside the assistant.
For a store, the change is that the first “visitor” is increasingly software reading your product data, not a person browsing your pages.
02
The two protocols
The Universal Commerce Protocol (UCP) is co-developed by Google with Shopify, Etsy, Wayfair, Target and Walmart, with others including Amazon, Microsoft and Stripe listed on its site. It covers the whole purchase: catalogue search, cart, account linking through OAuth, checkout and order management, with payments through AP2 payment mandates. The merchant remains the merchant of record and keeps the customer relationship.
The Agentic Commerce Protocol (ACP) is maintained by OpenAI and Stripe under the Apache 2.0 licence. It defines a checkout API and a delegated-payment API, and its latest stable version is dated 17 April 2026, with the specification still marked beta. OpenAI's merchant documentation starts with a structured product feed so ChatGPT can index a store's products accurately.
03
What is actually live
This is moving month by month. OpenAI launched in-chat checkout in ChatGPT in September 2025, and both protocols have published several versions since. Which assistants can complete a purchase, in which countries and for which stores, varies and changes often — check your platform's current documentation rather than a roundup article, including this one.
04
What a store should do now
None of this requires a rebuild. Most of it is product-data hygiene that also helps ordinary search.
- Complete, accurate product data: clear titles, attributes, identifiers such as GTINs, and true price and stock
- Structured data on product pages that matches what the page visibly says
- A current Google Merchant Center feed — Google names it as something to keep up to date for its AI features
- Shipping, returns and policies written as plain text an assistant can read
- Before building anything custom, check what your commerce platform already supports or has announced
05
Trade-offs
Selling through an assistant means less control over how your product is presented, and a customer who may never see your site. Fraud checks, tax, accounting and fulfilment still have to work for orders that arrive this way, so any integration should go through the same systems as your web checkout rather than around them.
For most stores the sensible position today is ready but not reshaped: product data an agent can trust, on a platform that will speak the protocols, without betting the site on a channel that is still settling.
Related services
Sources
- Universal Commerce Protocol — UCP
- Agentic Commerce Protocol (specification repository) — OpenAI and Stripe, GitHub
- Agentic commerce documentation — OpenAI
- AI features and your website — Google Search Central
Accessed 22 October 2026. Kramiva is not affiliated with these publishers.
Tell us what you are building.
A short brief gets you a real reply from a founder within one business day — an honest read on fit, scope, and budget.

