Course Overview
This 4-hour ILT course develops skills in building multi-agent solutions using ADK, MCP, A2A, Agent Engine, Cloud Run and Gemini Enterprise. You will build and deploy a realistic multi-agent system on Google Cloud with an integrated agent in Gemini Enterprise to solve a realistic use case related to customer engagement.
Who should attend
Application Developers, DevOps Engineers, ML Engineers and anyone deploying agentic applications on Google Cloud.
Prerequisites
- Completion of Deploy Multi-Agent Systems with Agent Development Kit and Agent Engine, or equivalent knowledge
- Strong familiarity with the business domains and processes being addressed
Course Objectives
- Create capable agents using Agent Development Kit
- Access remote tools using Model Context Protocol
- Implement sensitive data guardrails with Model Armor and Sensitive
- Data Protection templates
- Enable agentic system observability with Cloud Trace and Cloud Logging
- Build systems of agents the work together using the Agent2Agent (A2A) protocol
- Deploy agents to Cloud Run and Agent Engine
- Expose agent to users through Gemini Enterprise
Outline: Build Multi-Agent Systems Using Google's Agentic Stack (BMASGAS)
Module 1 - Getting Started
Topics
- Course introduction
- Pre-work discussion
- Introduce the challenge and its solution
Module 2 - Implement and test the data agent
Objectives
- Set up development environment
- Review base agent code
- Implement MCP tooling
- Enable logging and tracing
- Add Model Armor sanitization
- Create an A2A server
- Test locally and deploy to Cloud Run
Activities Lab: Implement and test the data agent
Module 3 - Implement and test the improvement agent
Objectives
- Review base agent code
- Add authentication tooling
- Build RemoteA2aAgent objects
- Create AgentTools
- Add tools to agent
- Test locally and deploy to Agent Engine
- Add agent to Gemini Enterprise
Module 4 - Complete the multi-agent solution
Objectives
- Complete and deploy second agent
- Review third agent code
- Implement MCP tooling
- Create an A2A server
- Deploy and test the entire solution
Activities Challenge Lab: Complete, deploy, and test the intervention agent