Agent-to-Agent Collaboration Architecture

Everyone is building agents. Almost nobody is building the protocol between them. That is the actual bottleneck. The moment your agent and mine need to get something done together, the interesting questions are not about intelligence. They are about plumbing: - How does my agent discover yours? - How does it know what yours can do, and what it must not touch? - How do they establish trust without routing every step back through us? These are protocol questions. The same kind HTTP and email had to answer before the internet could do anything useful. A2A (Agent2Agent) is where that layer is finally taking shape. Agent Cards for discovery. Structured tasks instead of freeform chat. Scoped auth. A shared vocabulary two agents from different vendors can actually speak. With MCP handling context, we have the beginnings of an open agent web. At Rizom we are building on top of this layer, not around it. A2A gives us the dial tone. Our job is what happens on the call: agents that carry their human's expertise and boundaries, and meet other agents as peers. "Have your agent call my agent" only works if there is a line to call on. https://yeehaa.io/presentations/have-your-agent-call-my-agent #A2A #MCP #AIAgents #Rizom