02 · Building agents¶
How agentic applications work under the hood, and how to build them with LangGraph, a widely used open-source framework that works with any model provider.
You'll be able to: explain the agent loop, design safe tools, build a LangGraph graph with checkpoints and human approval, pick a multi-agent pattern, and argue for a simpler design when an agent isn't needed.
| # | Note | Level | Time |
|---|---|---|---|
| 1 | The agent loop | Beginner | 10 min |
| 2 | Tool calling | Beginner | 12 min |
| 3 | LangGraph basics | Beginner | 15 min |
| 4 | Human-in-the-loop | Intermediate | 12 min |
| 5 | Multi-agent patterns | Intermediate | 12 min |
| 6 | When not to use an agent | Beginner | 10 min |
Tip
Code blocks marked # test: run work without an API key. CI runs them on every change, so they stay correct as libraries update.
Project: P3 · LangGraph research agent uses everything in this module, plus memory and evaluation. Do it after modules 03 to 06, or now if you're keen.