Google’s UCP and the next layer of agentic commerce: where Icecat’s MCP fits

Google has introduced the Universal Commerce Protocol (UCP), an open standard designed to help AI agents move beyond “shopping advice” into real transactions, spanning product discovery and offer checks through checkout, payment, and post-purchase flows.

What makes this announcement especially relevant for Icecat users is that UCP is built to work with existing agent standards, including Model Context Protocol (MCP) and Agent-to-Agent (A2A). In other words, the ecosystem is converging: agents need a shared way to understand products, talk to tools, communicate with other agents, and complete purchases safely.

So where does Icecat come in? And should Icecat Commerce support UCP? Let’s connect the dots, using insights from Icecat leaders who work closest to these building blocks.

UCP vs. MCP vs. A2A: who does what?

A helpful way to think about these protocols is by job:

• MCP helps agents access tools and context (for example: “fetch product specs,” “retrieve GTIN matches,” “pull structured attributes”). Icecat’s MCP server is designed exactly for this: making Icecat’s product knowledge available on demand inside AI workflows.
  
• A2A helps agents communicate and coordinate across systems.
  
• UCP helps agents execute commerce, pricing/availability checks, checkout, and payments, using a standardized “transaction layer.”
  

Marc Weehuizen, Managing Director of Icecat Commerce, summarizes the difference cleanly: “MCP and A2A enable agents to talk; UCP enables them to conduct business safely and efficiently (prices, stock, payment).” That framing matters. It explains that UCP isn’t replacing MCP, it’s extending the stack into the “now, buy it” moment.

Why structured product data becomes non-negotiable in agentic commerce

Agentic shopping only works if an AI can reliably compare and decide. And that starts with structured data, not marketing copy.

Marc puts it bluntly: “You can only compare products that are exactly the same. Without structured content, that would not be possible.” In practice, for AI systems to reliably compare and select products, product content must use consistent attributes, clean identifiers (such as GTINs), clear units, and category-specific specifications that align across brands.

This is also why Icecat has been investing in MCP: it’s a pathway for AI systems to query standardized product records and retrieve the “ground truth” they need for confident recommendations. Icecat’s MCP server brings large-scale structured product content into AI ecosystems, so agents can work with consistent, validated product knowledge rather than guessing from fragmented listings.

From “agent-readable” to “agent-actionable”: where Icecat Commerce enters

If MCP helps agents understand products, UCP helps agents purchase products. The bridge between those two is where many ecommerce teams will feel the pressure in 2026: they will need to connect product truth with commercial truth (price, stock, fulfillment rules, payment options).

That is exactly the territory Icecat Commerce is moving into.

Marc explains how Icecat’s Commerce stack prepares merchants and platforms for standards like UCP: “We can deliver standardized content, stock positions, pricing, and support the order and fulfillment process.” That’s the practical shift. It’s not only about enriching PDPs anymore; it’s about making product and commerce data usable in real time, in a format agents can consistently consume.

Guillaume Stritmatter, Icecat CTO, described this as a “two-part” flow: the browsing/decision phase and the transaction phase. UCP focuses on the transaction mechanics, but decisions still depend on reliable product information. His core point is simple: if an agent is going to choose on a shopper’s behalf, it must rely on official, consistent product data to reduce wrong purchases and returns, exactly what standardized Icecat content is built to provide.

Should Icecat Commerce support UCP?

Based on how Google is positioning UCP, compatible with MCP and A2A, and designed for broad adoption, the direction is hard to ignore. If UCP becomes a common “checkout language” for agents, then supporting it is less about chasing hype and more about protecting interoperability.

For Icecat users, the upside is clear:

• Less integration friction as agentic interfaces proliferate
  
• More control over “what the agent sees” (structured product truth, not scraped guesswork)
  
• A single environment for standardized content + commercial signals, which can serve real-time agent queries (pricing, availability, fulfillment)

Marc describes the end-state as a unified operating layer: “standardization, sales data, and management in one environment” that can support AI agents in real time. In that future, Icecat’s role expands from syndication to agentic commerce infrastructure.

Looking ahead: the ecosystem is converging

UCP is one more signal that ecommerce is shifting from pages and clicks to conversations and actions. People will still browse. But increasingly, agents will compare, decide, and transact.

In that world, winners won’t be the companies with the most content. They’ll be the ones with the most structured, interoperable, and trustworthy product truth, delivered in a way agents can use instantly. Icecat’s MCP work already aligns with that direction, and UCP may become the next logical layer for Icecat Commerce to support as the stack matures.

If agentic commerce is the new interface, then “being found” becomes “being understood.” And being understood becomes “being bought.”