A2A in 30 Minutes: Building Interoperable AI Agents

Until recently, AI agents were advancing in intelligence but remained isolated. Developers attempted to make them collaborate by repurposing protocols like MCP, which connects language models to external tools and data. However, MCP was never intended for agent-to-agent communication. This gap became clear as agents grew more complex and collaboration became essential. In April, Google and its partners announced the A2A protocol, an open standard specifically designed to let agents discover each other, communicate, and collaborate across ecosystems.

This talk introduces the A2A protocol and shows how it enables agents to work together in multi-agent, multi-modal, and even long-running scenarios. We will explain its core principles, such as open standards, secure communication, and capability discovery through Agent Cards. You will also see how agents negotiate tasks, exchange context and artifacts, and coordinate their user experiences. We will highlight how A2A complements protocols like MCP by enabling agents to orchestrate one another while still using MCP internally to access tools and data.

By the end of this session, you will understand what problems A2A solves, how it fits into the broader agent ecosystem, and how it improves on previous approaches. You will leave with a clear understanding of how to design agents that can truly interoperate and contribute to the emerging Internet of Agents.