Build a Multi-Agent System with ADK, MCP, and Gemini

Description

Codelab→ https://goo.gle/40jFpLw Video for developers → https://goo.gle/3ZJ6dor Video for data engineers → https://goo.gle/4tRO4CO Video for platform engineers → https://goo.gle/4qPZsMy In this episode of Agentverse, we move beyond simple chatbots to architect a production-grade multi-agent ecosystem. You'll learn how to decouple your tools using the Model Context Protocol (MCP), craft specialized agent workflows (Sequential, Parallel, Loop), and orchestrate them as reusable microservices with the Agent-to-Agent (A2A) protocol. We wrap up by enforcing governance using callbacks, plugins, and memory to keep your system under control. Chapters: 0:00 - Intro 02:30 - Decoupled tooling ecosystem - Setting up MCP servers 07:50 - Elemental familiars - Setting up workflow agents 13:33 - Using A2A for intelligent delegation 16:09 - Governance as code - Setting up interceptors 19:33 - Agent State Memory 21:21 - Outro 🔔 Subscribe to Google Cloud Tech → https://goo.gle/GoogleCloudTech #Gemini #GoogleCloud #Agentverse Speaker: Debi Cabrera Products Mentioned: Agent Development Kit, MCP, MCP Toolbox for Databases, Cloud Run