Agents That Talk Back: MCP, ACP, A2A, Oh my!
summary
Demos show a Python FastAPI MCP server exposing add/multiply/subtract/divide tools, an MCP inspector that executes those tools, and clients using standard IO and SSE with OAuth2; JSON-RPC carries prompts, tools and resources but has no built‑in auth so transport needs TLS and token auth. Three A2A agents (Alice/Bob/Carol) run a number‑guess workflow to prove agent discovery, streaming and callbacks work. An ACP demo implements a human‑approval password generation flow and demonstrates agent chaining and REST‑style interoperability.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 5:24Model Context Protocol
- 9:42Live MCP demo
- 16:09MCP security
- 20:23Agent-to-Agent protocol
- 27:51Building A2A agents
- 36:40Agent Communication Protocol
- 39:53Live ACP demo
- 45:00MCP vs A2A vs ACP
- 46:23Q&A
related talks
46mMCP and A2A Revolutionizing Interconnection and Collaboration of AI Agents
54mTour of Agent Protocols: MCP, A2A, AG-UI, A2UI
50mTour of Agent Protocols: MCP, A2A, AG-UI, A2UI
43mThe Protocol Wars: Clash of Titans
01h 23mA2A & MCP Workshop: Automating Business Processes with LLMs
36mPutting AI Agents to Work: A New Era of Open Source Connectivity with MCP
50mTutorial: Building Custom MCP Servers With FastMCP and Integrating AI Agents Using...
47mAgent Client Protocol for Java Developers