The short definition: AI agents that find, compare, negotiate and buy for someone, using machine-readable product data, checkout APIs and payment credentials the buyer approved. IBM and Salesforce describe it the same way: the agent acts for a consumer or a business, often without a person clicking through each step.
It stopped being a slide-deck idea in the past year. OpenAI and Stripe, Google, Visa, Mastercard and the major commerce platforms all shipped working pieces, and Gartner predicts that by 2028, 90% of B2B buying will be handled through AI agents, with more than $15 trillion of spend. B2B sellers are not first in line, but they are in the line.
What an agent actually does when it shops
A buyer, or a purchasing team, tells an assistant what it needs. The agent reads product data, shortlists and compares, then either hands the person a link to buy or completes the purchase itself. The important difference from the chatbot on your own site: this agent works for the buyer, not for you, and it may never look at your pages the way a person does.
- Discovery. The agent reads product feeds, structured data and pages to build a shortlist. This is where most of the activity is today, and where incomplete product data quietly drops you out of the comparison.
- Delegated checkout. The agent completes the purchase inside the AI app, through a checkout API and a payment token the buyer approved.
- Autonomous buying. Agents buy within set rules with no person present, such as replenishment or machine-to-machine payments. Google's AP2 v0.2 and Mastercard's Agent Pay for Machines are built for this, and it is still early.
The protocols, in plain terms
Several open standards shipped in 2025 and 2026. They overlap more than they compete, and the platforms are adopting more than one.
- Agentic Commerce Protocol (ACP). Released by OpenAI and Stripe in September 2025 alongside Instant Checkout in ChatGPT, open source under Apache 2.0 and still labeled beta. Its April 2026 release added carts, a product feed API and an MCP binding. In March 2026 OpenAI said the first version of Instant Checkout was not flexible enough and moved ChatGPT's focus to product discovery, with merchants using their own checkout.
- Universal Commerce Protocol (UCP). Announced by Google in January 2026, co-developed with Shopify, Etsy, Wayfair, Target and Walmart, and running checkout in Google's AI Mode and Gemini for eligible US retailers. Amazon, Meta, Microsoft, Salesforce and Stripe joined its Tech Council in April 2026.
- Agent Payments Protocol (AP2). Google's payments layer, launched in September 2025 with more than 60 partners. It keeps signed records, called mandates, of what the buyer approved. Google handed it to the FIDO Alliance in April 2026.
- Model Context Protocol (MCP). Anthropic's open standard for connecting AI models to tools and data, released in November 2024 and now part of the Linux Foundation's Agentic AI Foundation. ACP and UCP can both run over it, and commerce platforms now ship storefront MCP servers.
- Card network programs. Visa's Trusted Agent Protocol, built with Cloudflare, and Mastercard Agent Pay let agents pay with tokenized cards and help merchants tell a trusted agent from a scraper.
If you are wondering which one to bet on: you do not have to. Your commerce platform implements the protocols. Your job is the data and the rules underneath them.
Where the commerce platforms stand
Status as of September 2026, from each vendor's own announcements and documentation.
- BigCommerce. Added Stripe's Agentic Commerce Suite in December 2025 and endorsed UCP in January 2026. In April 2026 it announced agent-ready product feeds to ChatGPT, Gemini, Copilot and Perplexity through Feedonomics, plus AI purchase-order automation in B2B Edition. Its Storefront MCP server is labeled beta and covers guest shopping; the B2B server is listed as coming soon.
- Shopify. Agentic Storefronts sell into ChatGPT, Copilot, Google AI Mode and Gemini, and every store has a Storefront MCP server.
- Salesforce Commerce Cloud. Announced support for ACP in October 2025 and for UCP in January 2026, and made its B2B Buyer Agent generally available in June 2026.
- Adobe Commerce. Committed to UCP and ACP in February 2026 and introduced a Commerce MCP server covering catalog, cart, pricing, inventory, checkout and orders in April 2026.
- Sana Commerce and DynamicWeb. We found no public agentic commerce or MCP announcement from either as of September 2026. We build on both, and we are watching for one.
Why B2B is harder, and not covered yet
Consumer agent checkout assumes one buyer, one list price and one card. B2B buying runs on contract pricing, purchase orders, payment terms, approval chains and punchout from procurement systems, and the protocols have barely started on that.
- UCP. The August 2026 release added fractional quantities, payment terms and split payments. Its release notes do not mention purchase orders, contract pricing or approvals, and a proposal in the UCP repository explicitly leaves B2B configurators with contract pricing for later.
- ACP. The only change labeled B2B so far is support for decimal quantities.
- Punchout. We found no cXML or punchout support in either protocol.
- The vendors agree. Adobe names complex pricing, multi-step approvals and procurement workflows as the hard part. BigCommerce says authenticated shopping with personalized pricing comes next. Salesforce's B2B Buyer Agent handles multiple carts and round-trip quoting.
Buyers will not wait for the standards. Forrester predicted that in 2026 one in five B2B sellers would have to negotiate quotes with buyers' AI agents. Those agents will read whatever you publish, so the practical question is whether what you publish is right.
What a B2B seller should do now
None of this requires picking a protocol. It requires data an agent can trust and rules an agent can follow.
- Make product data complete. Agents shortlist from feeds and structured data. ChatGPT's product feed spec requires an item ID, title, description, URL, brand, image, price and availability, and Google added Merchant Center attributes such as product FAQs and compatible accessories for conversational search. Missing specifications leave you out of the comparison. Our catalog enrichment article covers filling the gaps without inventing facts.
- Keep price and stock true to the ERP. An agent that quotes a stale price or promises stock you do not have creates a dispute, not a sale. The ERP stays the source of truth, and the storefront and feeds have to follow it. See real-time vs middleware ERP integration.
- Publish your terms where machines can read them. Minimum order quantities, lead times, shipping, returns and payment terms belong on crawlable pages and in structured data, not only in PDFs or behind a login.
- Expose clean APIs. Agents act through your platform's APIs and MCP server. On BigCommerce that means the Storefront MCP server and B2B Edition APIs; on a custom stack it means an MCP endpoint with scoped access. Our own open-core engine, Wardenclyffe, ships buyer and merchant MCP endpoints for this reason.
- Decide what an agent may do alone. Which orders can an agent place on a buyer's account, up to what value, and what needs a person to approve? Write it down before the first agent order arrives.
- Let the AI crawlers in, and watch them. Allow the search crawlers the assistants use, keep an llms.txt file current, and check your server logs. On our own site, requests identifying as ChatGPT-User fetched pages on every one of the 32 days to September 14, 2026.
Frequently asked questions
What is agentic commerce?
Buying and selling where AI agents act for a person or a business: they search, compare and can complete a purchase, using product data, checkout APIs and payment credentials the buyer approved.
What is the difference between ACP and UCP?
Both are open protocols for agent checkout. ACP comes from OpenAI and Stripe and grew out of Instant Checkout in ChatGPT. UCP comes from Google with Shopify, Etsy, Wayfair, Target and Walmart, and runs checkout in Google's AI Mode and Gemini. Both can run over MCP, and Adobe Commerce and Salesforce have committed to both.
Is agentic commerce ready for B2B?
Not yet. The protocols handle a buyer, a cart and a payment. Contract pricing, purchase orders, approvals and punchout are not covered in UCP or ACP, and BigCommerce lists its B2B MCP server as coming soon. Sellers can prepare now with accurate product data, pricing that matches the ERP, and clean APIs.
Do I need an MCP server for my store?
You need what sits underneath one first: accurate catalog, pricing and availability data behind an API. BigCommerce and Shopify now provide storefront MCP servers, and custom platforms can add one.
How big will agentic commerce get in B2B?
Gartner predicts that by 2028, 90% of B2B buying will be handled through AI agents, with more than $15 trillion of spend. Treat forecasts like that as a direction rather than a budget line.
Facts checked against the vendors' own announcements, specifications and documentation on September 14, 2026. This area changes monthly.