Why do some AI agent standards spread fast while others stall — is it winning a fight, or avoiding one?

This explores whether AI protocols get adopted by beating competitors or by having the best design. The corpus points to a third answer that makes the choice look less stark: the protocols that spread tend to avoid head-on competition. Instead they position themselves as a layer that works alongside the others. If you expected a standards war with one winner, the material here suggests that picture is often wrong.

The clearest case is agentic commerce. A shopper's AI agent now has a growing list of acronyms to deal with: ACP, AP2, UCP, Visa's TAP, and Mastercard's Web Bot Auth. You'd expect them to fight it out. Do agentic commerce protocols compete or complement each other? argues instead that they form a 'capability ladder'. Each one handles a different job: running the application, governance, helping agents find merchants, or proving an agent's identity. Each also absorbs a different level of fraud risk. Merchants stack them rather than pick one. On this reading, a protocol wins adoption by owning a specific risk layer that nobody else covers, not by having the most elegant spec.

Agent coordination standards show the same pattern at the infrastructure level. Should coordination protocols wrap existing systems or replace them? finds that coordination layers get adopted when they wrap and connect protocols that already exist, like MCP and DIDComm, rather than trying to replace them. The reason is practical: bridging lets value build up step by step without forcing everyone to rewrite their systems at once. A technically better replacement can still lose to a merely adequate wrapper that asks less of the people adopting it. In effect, how cheaply a protocol fits into existing systems ends up mattering more than how good its specification is.

Where real competition does happen, it is less about protocol against protocol and more about platform architecture. Will agentic commerce platforms adopt open or closed models? compares the coming split to payment networks. Visa-style open systems let many agents take part, while closed systems require a platform's own agent, the way American Express ran its own network end to end. That decision, more than any spec detail, shapes which standards get adopted, how concentrated the market becomes, and who controls the full stack. So the market fight is real, but it plays out over whether the ecosystem stays open, and the protocols follow from that.

The corpus has a clear gap. It has little direct evidence on the specification side, such as standards bodies, formal ratification, or cases where a well-specified protocol won on technical merit. It also has no head-to-head adoption data. Treat the 'layer, don't compete' pattern as a strong argument from a few sources, not a settled finding. The useful takeaway: when you see a new agent protocol announced, the better question is usually not 'will it beat the others?' but 'which layer does it claim, and what does it wrap?'