Guides/Getting Started

Claude Code for Beginners

Claude Code is a CLI that brings an AI pair programmer directly into your terminal. This guide gets you from zero to productive in under 30 minutes.

Beginner FriendlyTerminalFree to Start

What is Claude Code?

Claude Code is an official CLI tool from Anthropic that runs Claude — the AI — inside your terminal. Unlike chat interfaces, Claude Code can read your files, edit your codebase,run commands, and remember context across a working session.

It is not a code autocomplete tool. It is a full peer that understands your project architecture, follows your conventions, and executes multi-step tasks autonomously.

Files

Reads, writes, and diffs any file in your project

Commands

Runs shell commands, installs packages, runs tests

Memory

Retains full context across a session via CLAUDE.md

Installation

Install via npm

# Install globally (requires Node.js 18+)

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

Authenticate

Run claude for the first time. You will be prompted to log in with your Anthropic account or enter an API key.

$ claude

Welcome to Claude Code. Authenticate to get started...

Navigate to Your Project

Always run Claude Code from your project root. It uses the current directory to discover files and context.

$ cd my-project && claude

First Commands to Know

Claude Code has slash commands for session management. Type any of these at the prompt:

/helpShow available commands and keyboard shortcuts

/statusShow current session context and memory state

/clearClear the conversation and start fresh

/compactCompress context when the window gets full

/modelSwitch between Claude models mid-session

/costShow token usage and estimated cost for the session

Understanding the Conversation Model

Claude Code works like a conversation, not a command queue. Context builds up across your session — what you said 10 messages ago still influences the current answer.

Session starts fresh

Every new claude invocation starts with an empty context window.

CLAUDE.md auto-loads

If a CLAUDE.md exists in your project root, it is injected automatically at the start.

Context accumulates

Every file read, command output, and message is added to the window. Rich context = better responses.

Context has a limit

Run /compact periodically on long sessions to compress history and free space.

Tips for Effective Prompting

The quality of your prompt directly determines the quality of the output. Here are the highest-leverage habits:

Be specific about scope

Vague

"Fix the bug in my code"

Specific

"Fix the null check error on line 42 of src/lib/auth.ts — it crashes when the user cookie is missing"

Provide context upfront

Vague

"Add a database table"

Specific

"Add a `notifications` table to src/lib/db.ts using the same UUID pattern as the existing `users` table"

State the outcome, not the method

Vague

"Use useEffect to fetch data"

Specific

"Show users their recent orders on the dashboard page — load from /api/orders"

Common Mistakes to Avoid

Asking too broadly

Break large tasks into specific, scoped steps. One thing at a time.

Skipping the CLAUDE.md

A well-written CLAUDE.md eliminates most repeated instructions. Write it once.

Not reading the output

Claude shows what it changed. Review every diff before accepting.

Fighting the conversation model

Build up context across a session. Don't restart for every question.

Ignoring /compact warnings

Run /compact when prompted. Bloated context degrades output quality fast.

No .env.local file

Always store secrets in environment variables. Never paste API keys into the chat.

Ready to Go Deeper?

The AI Starter Package includes a pre-configured CLAUDE.md, 143 skills, agent setup, and a full operating system for Claude Code — ready to deploy in minutes.

Get the AI Starter Package Write Your CLAUDE.md

What is Claude Code?

It is not a code autocomplete tool. It is a full peer that understands your project architecture, follows your conventions, and executes multi-step tasks autonomously.

Files

Reads, writes, and diffs any file in your project

Commands

Runs shell commands, installs packages, runs tests

Memory

Retains full context across a session via CLAUDE.md

Installation

Install via npm

# Install globally (requires Node.js 18+)

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

Authenticate

Run claude for the first time. You will be prompted to log in with your Anthropic account or enter an API key.

$ claude

Welcome to Claude Code. Authenticate to get started...

Navigate to Your Project

Always run Claude Code from your project root. It uses the current directory to discover files and context.

$ cd my-project && claude

First Commands to Know

Claude Code has slash commands for session management. Type any of these at the prompt:

/helpShow available commands and keyboard shortcuts

/statusShow current session context and memory state

/clearClear the conversation and start fresh

/compactCompress context when the window gets full

/modelSwitch between Claude models mid-session

/costShow token usage and estimated cost for the session

Understanding the Conversation Model

Claude Code works like a conversation, not a command queue. Context builds up across your session — what you said 10 messages ago still influences the current answer.

Session starts fresh

Every new claude invocation starts with an empty context window.

CLAUDE.md auto-loads

If a CLAUDE.md exists in your project root, it is injected automatically at the start.

Context accumulates

Every file read, command output, and message is added to the window. Rich context = better responses.

Context has a limit

Run /compact periodically on long sessions to compress history and free space.

Tips for Effective Prompting

The quality of your prompt directly determines the quality of the output. Here are the highest-leverage habits:

Be specific about scope

Vague

"Fix the bug in my code"

Specific

"Fix the null check error on line 42 of src/lib/auth.ts — it crashes when the user cookie is missing"

Provide context upfront

Vague

"Add a database table"

Specific

"Add a `notifications` table to src/lib/db.ts using the same UUID pattern as the existing `users` table"

State the outcome, not the method

Vague

"Use useEffect to fetch data"

Specific

"Show users their recent orders on the dashboard page — load from /api/orders"

Common Mistakes to Avoid

Asking too broadly

Break large tasks into specific, scoped steps. One thing at a time.

Skipping the CLAUDE.md

A well-written CLAUDE.md eliminates most repeated instructions. Write it once.

Not reading the output

Claude shows what it changed. Review every diff before accepting.

Fighting the conversation model

Build up context across a session. Don't restart for every question.

Ignoring /compact warnings

Run /compact when prompted. Bloated context degrades output quality fast.

No .env.local file

Always store secrets in environment variables. Never paste API keys into the chat.