Claude Code in depth

What you'll see in this lesson

1. Install Claude Code in two minutes

   Claude Code is a coding agent that works in your terminal; it installs with one command and needs a paid plan or the API.

2. Permission modes: how much it decides on its own

   Shift+Tab switches modes: from Manual, which asks before acting, to auto, where a classifier checks for you.

3. Plan mode: explore, plan, then code

   In plan mode Claude reads the code and proposes a plan without touching anything; it only edits once you approve.

4. Context, effort and usage under control

   Four commands tell you how much context you've used, compress it, adjust how much the model reasons and show what you've spent.

5. A short CLAUDE.md that actually gets followed

   CLAUDE.md loads in every session: include only what Claude can't guess and keep it under 200 lines.

6. Hooks: rules that are always followed

   A hook is a command Claude Code runs automatically at a specific moment, no matter what the model decides.

7. Subagents: delegate without filling your context

   A subagent works in its own context window and returns only the summary, without cluttering your conversation.

8. Skills, plugins and MCP: extend Claude Code

   Skills add procedures, MCP connects external services and plugins bundle it all so you can install it in one go.

9. Summary: the workflow of top-performing teams

   Plan before changing, control your context, automate the rules and give Claude a way to check its work.