Claude Code: Beginner

Beginner

Claude Code is Anthropic’s agentic coding assistant that runs in your terminal. It reads and edits files in your project, runs commands, and works through multi-step tasks, all from a chat prompt. This guide gets you from zero to a productive first session.

New to AI tools in general? Skim how AI coding tools work first. Everything here builds on it.

What you need

• A terminal (macOS Terminal, iTerm, Windows Terminal, or any Linux shell)
• A Claude account (Pro/Max subscription or an Anthropic API key)
• Node.js 22+ only if you install via npm (node --version to check); the native installer needs nothing

Install

1. Install the CLI:

   Native installer

   curl -fsSL https://claude.ai/install.sh | bash

   npm

   npm install -g @anthropic-ai/claude-code

   Homebrew

   brew install --cask claude-code

   These commands are current as of mid-2026; if one fails, check the official setup guide rather than fighting it.

2. Go to a project you want to work on and start it:

   cd my-project
   claude

3. On first run, follow the login prompt to authenticate with your Claude account.

Tip

Always start Claude Code from the root of your project. It uses the current directory as its workspace, so this determines which files it can see and edit.

Your first session

Once Claude Code is running, just talk to it in plain English. A few good first prompts:

What does this project do? Give me a short overview.

Find where the homepage title is set and change it to "Hello".

Claude will explore the files, show you what it plans to change, and ask for approval before editing or running commands. That explore-propose-approve loop is the agentic loop (see How AI coding tools work).

The commands you’ll actually use

Command	What it does
/help	List all available commands
/clear	Start a fresh conversation (clears the context window)
/rewind (or Esc twice)	Roll back to a checkpoint: code, conversation, or both
Esc	Stop Claude mid-action
!<command>	Run a shell command yourself, output goes into the chat

Caution

Claude asks before editing files or running commands. Read what it proposes before approving, especially anything that deletes files or touches git.

Three habits that make it work

1. Be specific. “Fix the bug” is weak. “The login form doesn’t show errors when the password is wrong. Fix it.” gives Claude something to work with.
2. Work in small steps. Ask for one change, check it, then continue. Don’t request a whole feature in one prompt yet.
3. Use /clear between unrelated tasks. A focused conversation produces better results than one giant thread.

You’re done when…

• ✅ You can start Claude Code in a project
• ✅ You’ve made and approved at least one file change
• ✅ You know how to rewind a change and how to clear context

Next step

Ready for real workflows: context files, planning, and iterating on bigger changes?

Claude Code: Intermediate Context files, plan mode, and @-mentions. The everyday workflow that turns Claude Code from a toy into a tool.