The Alphabet Soup of Agentic Standards: A Vocab Playbook for Merchants and PSPs

The biggest mistake merchants could make in agentic commerce may be waiting for one protocol to win. There probably won’t be one.

The emerging infrastructure for artificial intelligence (AI)-driven shopping is beginning to look less like the early browser wars and more like the internet stack itself, comprised of multiple protocols performing different jobs simultaneously. Google’s developer guide describes an ecosystem crowded with acronyms including MCP, A2A, UCP and AP2. Add OpenAI and Stripe’s Agentic Commerce Protocol (ACP), Visa’s Trusted Agent Protocol (TAP), Mastercard’s Agent Pay and emerging machine-payment standards, and the merchant technology roadmap starts resembling alphabet soup.

But what looks from the outside like a standards war may actually be the early construction of a layered architecture.

One standard may help an AI agent discover inventory. Another lets agents communicate. Another records what a consumer authorized. Another establishes that the agent knocking on a merchant’s digital door is legitimate. Yet another determines how payment credentials move.

The easiest way through the alphabet soup is to stop treating every protocol as a competitor.

Read more: Meta’s Agentic Strategy Is Coming for the Main Street Back Office 

Start With What the Acronyms Actually Do

Anthropic’s Model Context Protocol (MCP) connects AI systems with external tools and data. An agent might use it to access inventory, pricing or an internal merchant application programming interface (API).

Google’s Agent2Agent (A2A) addresses communication between agents themselves, allowing independently developed agents to discover capabilities and coordinate tasks.

The shorthand: MCP helps an agent use things. A2A helps agents talk to one another.

Neither necessarily tells an agent how to buy something. That’s where the commerce protocols enter.

The PYMNTS Intelligence report “Global Digital Shopping Index: The Agentic Commerce Deep Dive“ found that 56% of consumers will let an agent search and compare products on their behalf, and 35% will grant an agent access to their saved payment credentials.

OpenAI and Stripe’s Agentic Commerce Protocol (ACP) provides infrastructure connecting merchants with consumers shopping through AI interfaces, including product discovery and checkout.

Google and Shopify’s Universal Commerce Protocol (UCP) reaches across a broader commerce workflow, including identity, carts, checkout, discounts, loyalty and order management.

See also: Agentic AI Could Make Net 30 Obsolete 

What to watch: It is tempting to frame ACP and UCP as an OpenAI-versus-Google standards battle. But even here, a more pressing question for merchants may be whether they can expose the same catalog, inventory, pricing and checkout systems to multiple AI platforms. A merchant that cannot communicate with a major AI shopping platform could face the agentic equivalent of disappearing from search.

“We don’t see [agentic] so much as a new channel, as much as a new interface for interacting and engaging with commerce,” Sabrina Tharani, senior vice president, Global Fintech Programs at Mastercard, told PYMNTS this month.

Making products discoverable across that new interface is only half the problem. AI agents can increasingly operate without a consumer approving every individual step. That creates an authorization problem traditional eCommerce was not designed to solve.

Google’s Agent Payments Protocol (AP2) uses cryptographically signed mandates to establish what a consumer instructed an agent to do. A consumer could theoretically tell an agent to buy a certain product only below a specified price. The mandate provides evidence of those instructions when the transaction eventually occurs.

Related: AI Agents Push CFOs to Rethink Business Payments

Why it matters: Payments infrastructure traditionally proves that money moved. Agentic infrastructure now needs to prove why a machine was entitled to move it. That distinction is becoming important for disputes, fraud, liability and delegated purchasing.

When Machines Become Customers

Then comes trust. Visa’s Trusted Agent Protocol (TAP) helps agents provide verifiable information about their identity and intent to merchants. Mastercard’s Agent Pay similarly extends payment and tokenization infrastructure into agent-mediated transactions.

That flips decades of internet security on its head. Merchants historically built systems to identify and block bots. Now they need to distinguish bad bots from bots carrying legitimate customers.

Findings in the September edition of the Payments Innovation Tracker® Series, a PYMNTS Intelligence report done in collaboration with Paymentology, highlight that the industry is confronting a shift from isolated fraudulent transactions toward persistent identity-based attacks that can begin long before a payment is initiated, thanks to advances in artificial intelligence.

You may like: CFOs Turn to AI Harnesses as Agentic Capabilities Scale

Of course, the stack doesn’t stop with consumer commerce.

Stripe and Tempo’s Machine Payments Protocol (MPP) targets machine-to-machine transactions such as API calls, recurring digital services and micropayments. An agent can request a digital resource, receive a payment request and pay programmatically.

That suggests agentic payments could develop into two overlapping markets: agents buying for humans and machines buying from machines. And for merchants, across even one market alone, supporting every protocol independently would quickly become untenable.

Merchants generally don’t want to maintain separate integrations for every AI platform, agent framework, authorization standard and payment network. They want their catalog exposed, their checkout accessible and their payments secure regardless of which agent brings the customer.

That creates an opening for payment service providers (PSPs), commerce platforms and gateways.

The PYMNTS Intelligence report “Tech on Tech: How the Technology Sector Is Powering Agentic AI Adoption“ found a widening agentic readiness gap between tech companies and firms in goods and services, with 75% of tech firms reporting they were extremely familiar with agentic AI, versus 33% of goods firms and 38% of services firms.

For all PYMNTS digital transformation and B2B coverage, subscribe to the daily digital transformation and B2B newsletters.