Week 3: Multi-Agent Architectures & Protocols
This week focuses on building collaborative AI systems where multiple specialized agents work together. You'll learn industry-standard tools like CrewAI and protocols like MCP for agent interoperability.
Learning Objectives
By the end of this week, you will:
- ✅ Understand multi-agent system fundamentals and design patterns
- ✅ Build teams with CrewAI (Agent, Task, Crew, Process)
- ✅ Implement Hierarchical vs Sequential agent processes
- ✅ Use Model Context Protocol (MCP) for standardized tool connections
- ✅ Enable Agent-to-Agent (A2A) communication
Sessions
The Big Question
Can a team of gpt-4o-mini agents outperform a single gpt-4o genius?
Spoiler: Yes! Specialized agents with focused contexts often produce better results than one model trying to do everything.
Architecture Preview
Key Concepts
| Concept | Description | Framework |
|---|---|---|
| Role Playing | Agents with rich personas and backstories | CrewAI |
| Sequential Process | Tasks executed one after another | CrewAI |
| Hierarchical Process | Manager delegates to specialized agents | CrewAI |
| MCP Server | Exposes Resources, Prompts, and Tools | MCP |
| A2A | Agent-to-Agent communication | Custom |
Notebooks
| Session | English | Korean |
|---|---|---|
| 01. Router Agents | 📘 Notebook (opens in a new tab) | 📘 노트북 (opens in a new tab) |
| 02. Handoff Pattern | 📘 Notebook (opens in a new tab) | 📘 노트북 (opens in a new tab) |
| 03. Supervisor | 📘 Notebook (opens in a new tab) | 📘 노트북 (opens in a new tab) |
| 04. MCP Protocol | 📘 Notebook (opens in a new tab) | 📘 노트북 (opens in a new tab) |
| 05. CrewAI | 📘 Notebook (opens in a new tab) | 📘 노트북 (opens in a new tab) |
| Weekend Project | 📘 Project Page | 📘 프로젝트 페이지 |