A2A & MCP Workshop: Automating Business Processes with LLMs
summary
Shows a working webhook-to-agent pipeline that spawns A2A agents on incoming webhooks, uses Google's Agent-to-Agent protocol plus Zapier MCP and a Gemini API key, and successfully posts a Slack message and creates a GitHub issue during the demo. Host agent orchestrates parallel subagents that keep large tool results out of the main context, reducing costs with prompt caching and an observed optimal cache around 30,000 tokens. Notes practical limits: about a five-agent limit on subcalls and roughly 10,000 available MCP tools.
AI-written summary · about
chapters (10)
- 0:00Introduction to A2A and MCP
- 2:41Understanding protocols
- 4:48When to use A2A vs MCP
- 9:07Motivation for the protocols
- 10:54Repository and setup
- 15:08Agent architecture overview
- 20:26Live system demonstration
- 32:32Context management and cost
- 38:46Code and implementation details
- 44:44Q&A and final discussion
related talks
46mMCP and A2A Revolutionizing Interconnection and Collaboration of AI Agents
36mPutting AI Agents to Work: A New Era of Open Source Connectivity with MCP
50mAgents That Talk Back: MCP, ACP, A2A, Oh my!
01h 03mAI Coding Agents and how to code them
40mAgents in Action: Advancing Open Source AI Missions
01h 28mAI Coding Agents and how to code them
40mBuilding Agents with Agent Bricks and MCP
26mArchitecting Enterprise Agents Ecosystem: Standardizing APIs with MCP and Orch…