Agent Communication Protocols

Agent communication protocols specify how agents find peers, structure messages, manage multi-turn interactions, and coordinate work across tools and organizations. Approaches range from interoperability standards such as MCP and A2A to formal methods like session types and process calculi, alongside learned, compressed, or semantically routed messages. Research tests these designs in settings including software engineering, robotics, scientific workflows, and multi-agent reinforcement learning, with attention to latency, communication cost, security, and failure recovery. A central tension is that transport and message formats are becoming better defined, while shared meaning, verification, governance, and the right balance between protocol complexity and application-level flexibility remain open challenges.