When AI Agents Work Together: A2A, MCP, and Connected AI | Udaiappa Ramachandran

AI agents are becoming more capable, but the bigger opportunity is when they can work together. In this session, I walk through Agent2Agent (A2A), the open protocol for communication between independent AI agents, and how it fits alongside the Model Context Protocol (MCP). We look at the practical difference between MCP and A2A, how Agent Cards and discovery work, how agents communicate using messages, tasks, streaming, and long-running operations, and where Microsoft Agent Framework fits into the picture. I also walk through practical .NET and Python examples showing how to expose an agent through A2A, call a remote agent, work with streaming and background tasks, and combine A2A with MCP in real-world architectures. The session covers A2A vs. MCP, Agent Cards and discovery, messages and tasks, streaming and push notifications, Microsoft Agent Framework integration, exposing and consuming A2A agents, multi-agent delegation patterns, cross-framework collaboration, and how MCP and A2A can work together. The goal is to make A2A practical and easier to understand, especially if you are already building agents and wondering how independent agents, frameworks, and vendors can collaborate without creating another layer of custom integration code. Resources and samples are included at the end of the presentation. Connect with me: Udaiappa Ramachandran (Udai) https://udai.io   / udair