The more interesting question to me is how trading agents find each other in the first place. You can define a standard for how agents negotiate trades, but that assumes they already know which counterparty agents exist and what they offer.

Right now agent discovery is pretty fragmented. The official MCP Registry lists servers but only for MCP. Google's A2A has its own agent cards format. The IETF has at least 8 competing drafts for agent discovery (ARDP, AID, AINS, etc.), and the original agents.txt draft expires in 4 days with no clear successor.

A trading protocol like this only works at scale if there's a cross-protocol discovery layer underneath it. Otherwise you end up with agents that can speak the same trading language but have no way to find counterparties.