How to Use Claude Code: A Complete Step-by-Step Tutorial

Claude Code installs as a global CLI tool. Pick the method that matches your OS. Native installs auto-update — no manual version management required.

macOS & Linux

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

Windows — PowerShell

irm https://claude.ai/install.ps1 | iex

Windows — Command Prompt

curl -fsSL https://claude.ai/install.cmd -o install.cmd && install.cmd && del install.cmd

npm (any OS)

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

Verify the install by running claude --version. If you see a version number, you're ready for Step 2.

Run claude for the first time. It will prompt you to authenticate before doing anything else.

$ claude
No active session found. Opening login...

A browser window opens to complete OAuth. Once you authenticate, your credentials are stored locally and you won't be asked again.

Supported account types

• Claude Pro / Max / Team / Enterprise — claude.ai subscription
• Anthropic Console — pay-per-token API access at console.anthropic.com
• Amazon Bedrock — Claude hosted in AWS
• Google Cloud Vertex AI — Claude hosted in GCP
• Microsoft Azure AI Foundry — Claude hosted in Azure

Always navigate to your project directory first. Claude Code reads your files from the working directory — context is everything.

$ cd /path/to/your/project
$ claude

  Claude Code v1.4.0
  Model: claude-sonnet-4-5
  Directory: /path/to/your/project

  Type /help for commands, /exit to quit

claude>

The startup screen confirms three things: the version, the model in use, and the working directory. If the directory is wrong, exit and cd to the right place.

Essential session commands

• /help — list all available slash commands
• /resume — continue a previous session with its full message history
• /clear — wipe context and start fresh within the same session
• /exit — quit Claude Code

Start by having Claude understand your project. It reads files automatically — no manual copy-pasting, no context loading. Just ask.

• what does this project do?
• what technologies does this project use?
• where is the main entry point?
• explain the folder structure

Claude will scan package.json, README.md, config files, and source directories to build an accurate answer. It tells you which files it reads, so you can see exactly what it's working from.

Example exchange:

claude> what technologies does this project use?

Reading package.json... src/index.ts... tsconfig.json...

This is a TypeScript Node.js API using:
• Express 4.18 for HTTP routing
• Prisma 5.x as the ORM, connected to PostgreSQL
• Zod for request validation
• Jest for testing (see jest.config.ts)
• Docker Compose for local DB setup (docker-compose.yml)

Entry point is src/index.ts, which bootstraps Express and
registers routes from src/routes/.

Ask Claude to write something small and concrete. The "hello world" pattern is a reliable way to see the full approval workflow in action.

• add a hello world function to the main file

Here's what happens next — Claude runs through a three-step loop before touching anything:

1. Finds the file — locates src/index.ts (or wherever main lives in your stack)
2. Shows the diff — displays exactly what it intends to add or change
3. Asks for approval — waits for your "yes" before writing anything

Accept all mode

If you trust Claude for a session and want to skip approval on each edit, type accept all or press the corresponding key when prompted. This applies for the duration of the session only.

Claude Code has native git awareness. Ask in plain English — it translates to the correct git commands without you needing to remember flags.

• what files have I changed?
• commit my changes with a descriptive message
• create a new branch called feature/quickstart
• show me the last 5 commits
• help me resolve merge conflicts

Example — commit workflow:

claude> commit my changes with a descriptive message

Running: git status
  M  src/index.ts
  M  src/routes/users.ts

Running: git diff (reading changes...)

Proposed commit message:
  "feat: add helloWorld function and update server startup log"

Shall I commit with this message? (y/n)

This is Claude Code's core use case. Describe what's broken or what you want, and Claude locates the relevant code, understands the context, and implements a solution.

• add input validation to the user registration form
• there's a bug where users can submit empty forms — fix it

What Claude does internally:

1. Locates the registration form component
2. Reads the current validation logic (or notes the absence of it)
3. Checks existing patterns in your codebase (Zod, Yup, native HTML, etc.)
4. Implements validation consistent with your stack
5. Runs existing tests to verify nothing broke
6. Shows a diff for your approval

For bugs, describe the symptom rather than the cause. "Users can submit empty forms" is more useful than "the validation function returns true incorrectly" — Claude will find the cause.

Once you're comfortable with the basics, these are the four workflows you'll use daily:

Two categories: shell commands you run before starting a session, and slash commands you use inside one.

Shell commands

Session commands (inside Claude)

The basics get you 80% of the value. These features unlock the rest — especially for teams and complex projects.

Available interfaces

Claude Code isn't only a terminal tool. The same agent runs across multiple surfaces:

For CI/CD use cases, claude -p "..." in non-interactive mode integrates cleanly into pipelines. See the official quickstart docs for GitHub Actions YAML examples.

AI Coding Agent Myths vs. Truths

Recommended Developer Resources