Signal Three competing agent interoperability protocols are named with their stated limits

Summary

Deloitte describes three emerging protocols that let specialised agents work together in orchestrated workflows. Anthropic's Model Context Protocol standardises how AI systems connect to data sources and tools, offering agents a universal way into enterprise resources, though it struggles with complex security requirements and legacy integration. Google's Agent-to-Agent Protocol allows agents on different platforms to communicate directly, covering discovery, task delegation and collaborative workflows. The open Agent Communication Protocol lets agents interact through a RESTful interface regardless of where they were built, but may be limited in how many agents can coordinate in one network and in integrating with existing tools. The report sees these standards as the base of a microservices approach to AI, in which many small agents run close to their data and instructions. Benefits include easier debugging and maintenance, scalable orchestration and platform flexibility.