Tracks · Learn Claude

What is Claude Code, and how do you use it?

Claude Code is Claude running as an agent against your repository — a loop that decides, acts, observes and decides again, until a stopping condition you defined. This track covers orchestrator versus router, when fanning out subagents costs coherence, run modes from interactive to headless, hooks as deterministic gates, carrying state across steps, and when to stop. Assumes Tools & MCP.

What you learn in this track

The agent loop

An agent is a loop: decide, act, observe, decide again — until a stopping condition you defined.

Orchestrator vs. router

A router picks one path; an orchestrator owns a plan. Choosing wrong shows up as agents that never finish.

Task decomposition & plan ownership

Somebody has to own the plan, and it should be the layer that can see the whole job.

Fan-out trade-offs

Fanning out subagents buys speed and costs coherence — the joins are where the work is.

Run modes: interactive, headless, background

Headless runs need stopping conditions that interactive runs get from you.

Hooks as guards

Hooks are deterministic gates: they run whether or not the model decided they should.

State & memory across steps

An agent that forgets between steps will redo work it already did, confidently.

Termination & failure recovery

Deciding when to stop is a design decision, not an emergency measure.

Guides in this track

• Claude subagents or multi agent systems — when worth it?
• Agentic AI: how do you stop an agent loop that won’t finish?

How the Socratic coach teaches this

Each module is a Socratic dialogue: the coach probes your reasoning with scenario questions, surfaces weak points, and generates flashcards and study notes from your own gaps. Mastery is tracked per concept, so you can see which ideas you actually handle and which ones still need work.

See where this track sits in the curriculum on the six tracks and why they are in this order, or open the ClaudeMaster.ai hub to start it.