Phase 3: Agent Architect
Goal: The learner moves from building applications to building agents — AI systems that can use tools, access external data, and take actions. This is where the AI stack becomes deeply practical.
What You Can DO After This Phase
Section titled “What You Can DO After This Phase”- Explain what an AI agent is and how it differs from a chatbot
- Configure Claude Code with CLAUDE.md, hooks, and custom settings
- Set up MCP servers to give AI access to external tools
- Understand memory and RAG at a practical level
- Design single-agent systems with clear tool boundaries
- Reason about what should be automated vs. what needs human judgment
Sections
Section titled “Sections”| Section | Topic |
|---|---|
| 3.1 — What Are Agents | The chatbot–agent spectrum, tool use, and Claude Code as an agent |
| 3.2 — The Agent Stack | Expanded 8-layer stack with instructions, tools, and memory layers |
| 3.3 — Harness Engineering | CLAUDE.md, settings, hooks, and persistent memory |
| 3.4 — MCP | The USB standard for connecting AI to external systems |
| 3.5 — Memory & RAG | File-based memory vs. retrieval-augmented generation |
| 3.6 — Agent Patterns | Four architectural patterns for how agents work |
| 3.7 — Vocabulary | 20 terms to know cold before Phase 4 |
Previous phase: Phase 2 — Builder’s Toolkit | Next phase: Phase 4 — Orchestrator