Most of the agent protocol conversation has centered on MCP for tool access and A2A for agent-to-agent coordination. A newer, narrower entrant called X42 is aimed at a specific gap: what happens when an agent built on one platform needs to invoke an agent or capability running on somebody else's. That's a real operational problem as multi-agent systems move from demos to production. A sales agent might need to trigger a workflow at a partner's system. A personal agent might invoke a capability exposed by a third-party service. Each of those crossings raises the same three questions: who is this request really from, what is it allowed to do, and can anyone reconstruct what happened afterward. According to a recent industry rundown of the emerging agent-protocol stack, X42 tries to answer this by treating every cross-boundary invocation as something that has to carry its own proof. The protocol wraps inter-agent invocations in a permission model built from signed requests, scoped tokens, and execution logs meant to create a record of what was called, by whom, and under what authorization. That's a meaningfully different design point than the general-purpose coordination protocols it sits alongside. A2A handles discovery and task delegation between agents; MCP solves the tool-access problem for a single agent talking to its own environment. X42's proposition is that the risky part isn't discovery or tool-calling in isolation — it's the moment an agent's authority has to be trusted by a system it doesn't belong to, and that moment needs its own artifact trail rather than an inherited one. Whether X42 becomes a standard with real adoption or stays a niche pattern is an open question — it's explicitly framed as one of the newer, less mature entries in the stack, alongside AG-UI and A2UI, compared to the more established MCP and A2A. But the design instinct is worth noting on its own: as agents increasingly act across organizational boundaries rather than within a single vendor's walled garden, the industry is starting to treat every cross-boundary call as an event that needs its own signature and its own log, not just a shared protocol everyone happens to speak. The bet embedded in X42 is narrow but telling: coordination protocols get you interoperability, but interoperability without a built-in accountability trail just moves the trust problem somewhere less visible. Putting scoped tokens and execution logs directly into the invocation itself, rather than bolting them on at the infrastructure layer afterward, is a small but distinct architectural choice — one that treats every agent-to-agent handoff across a boundary as evidence-generating by default.