Claude Code Complete Guide: Installation, Setup, Commands & Best Practices

Claude Code Complete Guide

Claude Code is an AI-powered coding agent from Anthropic that can understand a codebase, edit files, run commands, work with Git, connect to external tools through MCP, and automate development workflows.

Unlike a traditional autocomplete tool, Claude Code can work across an entire repository and perform multi-step tasks such as investigating bugs, modifying multiple files, running tests, reviewing changes, and creating commits.

This guide covers everything you need to get started with Claude Code from installation to advanced workflows.

What You Need

Claude Code currently supports:

• macOS 13+
• Windows 10 1809+ / Windows Server 2019+
• Ubuntu 20.04+
• Debian 10+
• Alpine Linux 3.19+
• x64 and ARM64 processors
• At least 4 GB RAM
• Internet connection
• Bash, Zsh, PowerShell or CMD

On Windows, Claude Code can run natively or through WSL 2. Git for Windows is optional for native Windows, but installing it enables Claude Code's Bash tool through Git Bash.

1. Installing Claude Code

macOS, Linux or WSL

Open your terminal and run:

Windows PowerShell

Open PowerShell and run:

Windows CMD

Run:

Make sure you use the correct command for the terminal you're using.

PowerShell normally starts with:

CMD normally starts with:

The current official installation documentation recommends the native installer for most users.

2. Verify Installation

After installation:

You should see the installed Claude Code version.

For a deeper installation and configuration diagnostic:

claude doctor performs read-only diagnostics from the terminal and can identify installation, configuration and PATH problems.

3. Start Claude Code

Navigate into your project:

Then start Claude Code:

On the first launch, you'll be asked to authenticate.

You can also start Claude with an initial instruction:

For example:

4. Authentication

Claude Code can authenticate through supported Claude/Anthropic accounts and supported third-party providers.

You can also explicitly log in:

Check authentication:

Log out:

Claude Code can also use an ANTHROPIC_API_KEY environment variable.

5. Your First Claude Code Prompt

Instead of immediately asking Claude to modify code, start by asking it to understand the project.

Good first prompt:

Analyze this entire project.

Do not modify anything yet.

Explain:
1. The framework being used
2. The folder structure
3. The main entry points
4. How authentication works
5. How the database is connected
6. How API calls are organized
7. How the frontend communicates with the backend
8. How the project is built and deployed

This gives Claude context before making changes.

6. Give Claude a Task

Once Claude understands the project, you can ask it to implement something.

Example:

Add dark mode to the application.

First inspect the existing styling architecture.
Reuse the existing design tokens and components.
Do not introduce a new UI library.
After implementing it, run the relevant tests and report what changed.

Claude Code can inspect the repository, edit multiple files, execute commands and verify its work.

7. CLAUDE.md

One of the most useful Claude Code features is CLAUDE.md.

Create one in your project's root:

You can use it to provide persistent project instructions.

Example:

# Project Instructions

## Stack

- Next.js
- TypeScript
- Tailwind CSS
- PostgreSQL

## Rules

- Use TypeScript.
- Do not use `any` unless absolutely necessary.
- Reuse existing components.
- Do not introduce dependencies without asking.
- Keep components small and reusable.
- Run tests after significant changes.

## Git

- Never push directly to main.
- Use descriptive commit messages.
- Do not modify unrelated files.

Claude Code reads CLAUDE.md at the beginning of sessions and can use it for coding standards, architecture decisions and project-specific instructions.

8. Initialize CLAUDE.md Automatically

Inside Claude Code:

This initializes a project guide and can create a starter CLAUDE.md.

9. Most Important Claude Code Commands

Start Claude

Start an interactive session.

Start with a prompt

Non-interactive mode

Continue the previous conversation

or:

Resume a specific session

Update Claude Code

Check installation

Check authentication

Configure MCP

View background-session logs

Stop a background session

Attach to a background session

Start a cloud session

Select a model

Select effort

Enable Chrome integration

Run in print mode

Output JSON

Limit agentic turns

Set a budget limit

The official CLI reference contains the complete current command and flag reference, including newer commands for background agents, cloud sessions, MCP, plugins, remote control and more.

10. Important Slash Commands

Inside a Claude Code session, type / to open the command menu.

Some of the most useful commands are:

Claude Code's command availability can vary depending on platform, plan and environment, so the / menu inside your installation is the authoritative list for your session.

11. Useful Keyboard Shortcuts

12. Shell Mode

Claude Code can execute shell commands directly when you prefix a command with !.

Examples:

This allows you to run a command directly and add its output to the current Claude Code context.

13. Git Workflow

Claude Code works directly with Git.

For example:

Review my current Git changes.
Find any bugs.
Fix the issues you find.
Run the tests.
Then create a descriptive commit.

Claude Code can inspect diffs, create commits, work with branches and help prepare pull requests.

Always review generated changes before pushing them.

14. Plan Mode

For large changes, ask Claude to plan before modifying files.

Use:

Or:

This is useful when the change affects multiple parts of a codebase.

15. MCP

MCP stands for Model Context Protocol.

It allows Claude Code to connect to external tools and data sources.

Examples include:

• databases
• project management tools
• documentation
• GitHub
• Slack
• custom internal tools

You can manage MCP servers with:

Claude Code's documentation describes MCP as a way to connect external data sources and tools to Claude Code.

16. Background Agents

Claude Code can run work in the background.

Useful for:

• long-running tests
• builds
• research
• multiple development tasks
• parallel work

You can inspect background work with:

You can also start a background session from the CLI:

17. Code Review

Claude Code includes workflows for reviewing changes.

For example:

Security-focused review:

You can also ask:

Review the current diff.

Look specifically for:
- authentication vulnerabilities
- authorization issues
- SQL injection
- XSS
- exposed secrets
- insecure API endpoints
- sensitive information leaks

18. Common Prompt Pattern

A good Claude Code prompt usually contains:

CONTEXT
What project/file/system are we working on?

GOAL
What should change?

CONSTRAINTS
What should NOT change?

IMPLEMENTATION
What approach should be followed?

VALIDATION
How should the result be tested?

Example:

CONTEXT:
This is a Next.js TypeScript application using PostgreSQL.

GOAL:
Add password reset functionality.

CONSTRAINTS:
Do not change the existing authentication provider.
Reuse the existing email service.

IMPLEMENTATION:
First inspect the current authentication architecture.
Then implement the reset-token flow.

VALIDATION:
Run the relevant tests and verify the complete flow.
Do not modify unrelated files.

19. Safety Tips

Do not blindly approve every command Claude wants to run.

Be especially careful with:

database deletion commands, production deployments, credential changes and commands that modify large parts of a repository.

Claude Code provides permission modes and permission rules for controlling tool access.

Avoid using:

unless you fully understand the consequences and are operating in an appropriately isolated environment. The flag explicitly skips permission prompts.

20. Recommended First Workflow

For a new project:

1. Install Claude Code
2. Open your project
3. Run claude
4. Ask Claude to analyze the repository
5. Run /init
6. Review CLAUDE.md
7. Start with /plan for large changes
8. Implement one feature at a time
9. Review /diff
10. Run tests
11. Run /code-review
12. Commit the changes

Quick Reference

Installation

# macOS / Linux / WSL
curl -fsSL https://claude.ai/install.sh | bash

# Windows PowerShell
irm https://claude.ai/install.ps1 | iex

# Verify
claude --version

# Diagnostics
claude doctor

Starting

claude
claude "Analyze this project"
claude -p "Review this code"

Sessions

claude -c
claude -r "session-name"
claude --cloud "Fix this bug"

Authentication

claude auth login
claude auth status
claude auth logout

Development

/init
/plan
/diff
/review
/code-review
/security-review
/verify

Context

/compact
/context
/memory
/clear

Integrations

/mcp
/plugin
/skills

Background work

/tasks
/background
/fork
/subtask

Official Documentation

For the latest installation instructions and command reference, always check the official Claude Code documentation because commands and features can change between releases.