Skip to main content

What Claude Code Is

Beginner
What you'll learn
  • Explain what makes Claude Code agentic, not just a chat window
  • Picture the agentic loop: goal, action, permission, observe, repeat
  • Name the surfaces where Claude Code runs and how settings travel with you
  • Order the things you configure by leverage, starting with CLAUDE.md
  • Walk through the shape of a safe first session using Plan Mode

Claude Code is Anthropic's agentic coding tool. Unlike a chat window, it can actually do things in your project: read and edit files, run shell commands, search the codebase, and call external tools — all with your permission.

The mental model: an agentic loop​

This is the one idea that makes everything else make sense. You give an objective in plain language ("add tests for the auth module and fix what fails"). Claude plans, acts, observes the result, and repeats until the goal is met. You stay in control via permissions and Plan Mode.

Pro tip
  • The loop only advances on actions you allow. Nothing edits or runs without passing through that permission gate — which is exactly why the next sections matter.

Where you can run it​

The same Claude Code follows you across surfaces — it shares your settings, hooks, and permissions wherever you work.

  • Terminal (CLI) — the original surface; works in any shell.
  • IDE extensions — VS Code and JetBrains, with inline diffs.
  • Desktop and web — and it shares your settings, hooks, and permissions across surfaces.
  • Cloud sessions — kicked off from claude.ai, mobile, or claude --cloud, running on Anthropic's infrastructure by default; Team and Enterprise plans can point them at self-hosted runners inside their own network instead.

What you'll configure (in rough order of leverage)​

Think of this as a ladder: master the top rungs first, then layer on power features only when a real need appears.

Guided walkthrough1 of 5
  1. Persistent project instructions. Highest impact, lowest effort — start here.

Each rung links into its own lesson: CLAUDE.md, Plan Mode, Permissions, settings.json, Slash commands, hooks, skills, subagents, and MCP servers.

Your first session (the shape of it)​

Guided walkthrough1 of 5
  1. See the official docs for current commands.

Two commands worth memorizing from that first session:

Scaffold project instructions

/init

A safe, read-only first ask

Explain how routing works in this app.

For current install and authentication commands, see the official docs.

Pro tip
  • Start read-only. For your first real task, use Plan Mode — Claude investigates and shows you a plan without touching files. It's the safest way to build trust.

Key terms at a glance​

Claude Code vocabulary
Press Enter or Space to flip the card. Use the left and right arrow keys to move between cards.Term shown.
1 / 5

Check yourself

0/3
  1. What makes Claude Code different from a chat window?
  2. In the agentic loop, what happens right after Claude proposes an action?
  3. Which setup step is highest impact for the lowest effort?
Key takeaways
  • Claude Code is agentic: it acts in your project with your permission, not just chats.
  • The loop is goal, propose, allow, run, observe, repeat — you control it through permissions and Plan Mode.
  • It runs in the terminal, VS Code/JetBrains, and on desktop and web, sharing settings, hooks, and permissions across surfaces.
  • Configure by leverage: CLAUDE.md first, then Plan Mode, Permissions, settings.json, then power features.
  • Start a first session read-only in Plan Mode to build trust before letting edits run.

Next​