Agent Frameworks

About This Module

By now you've seen that an agent is, at heart, a loop: call the model, run a tool, feed back the result. You can write that loop yourself, but as agents grow you'll want help with state, control flow, retries, streaming, and human approval. Agent frameworks package those pieces so you don't rebuild them for every project.

This module looks at two different styles. LangGraph is a low-level orchestration framework: you describe your agent as a graph of steps with shared state, which gives you precise control over long-running, stateful workflows. The Claude Agent SDK is higher-level: it gives you a ready-made agent loop with built-in tools that you configure and extend. The first two lessons teach LangGraph, from beginner concepts to a longer project-based course, and the last one is a hands-on setup of the Claude Agent SDK.

As you watch, compare how much each approach decides for you and how much it leaves in your hands. A framework is worth adopting when it saves you real work, not just because it exists, and simple agents often need nothing more than the loop you already know.