The complete Claude guide: from beginner to power user

What exactly is Claude?

Claude is Anthropic’s AI assistant. It exists in two main forms:

Claude.ai - the conversational web interface (like ChatGPT)
Claude Code - the command-line tool for developers

This guide covers both. We start with Claude.ai for the basics, then level up to Claude Code for advanced users.

Part 1: Claude.ai, the fundamentals

Accessing Claude

Head to claude.ai. Three tiers:

Free: access to Claude Sonnet, usage limits
Pro ($20/month): Claude Opus, more messages, projects
Team ($30/month/person): team collaboration, admin features

The models

Anthropic offers three model tiers:

Haiku: fast and lightweight, ideal for simple tasks
Sonnet: balanced performance/speed, the default choice
Opus: the most powerful, for complex tasks

Effective conversations

The key to using Claude well:

Be specific. Not “help me with my resume,” but “rewrite the experience section of my resume emphasizing quantified results rather than responsibilities.”

Give context. Paste the document, code, or data directly into the conversation. Claude handles long contexts (200K tokens).

Iterate. The first response isn’t always perfect. Say what’s wrong, ask for adjustments. It’s a conversation, not a form.

Artifacts

Artifacts are interactive content Claude generates in a side panel:

Executable code
Formatted documents
Visualizations (charts, diagrams)
Interactive HTML pages

Projects

Projects (Pro/Team) let you:

Group conversations by theme
Add custom instructions applied to every message
Upload reference files accessible across all project conversations

Custom instructions

In settings, you can define instructions that apply to all conversations:

Your preferred communication style
Your area of expertise
Your preferred response format

Part 2: Claude Code, leveling up

Claude Code transforms Claude into a development teammate. It runs in your terminal with access to your filesystem, Git, and all your tools.

Installation

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

Then in any project:

claude

For a detailed installation guide, see our installation guide.

The CLAUDE.md file

This is the most important file. Placed at your project root, it gives Claude Code permanent context:

# My Project

## Stack
- Astro v6, Tailwind CSS v4, TypeScript
- Deployed on Vercel

## Conventions
- Components in src/components/
- Pages in src/pages/
- Always use absolute imports

Full guide: CLAUDE.md best practices

Essential commands

Command	Description
`claude`	Start an interactive session
`claude "do X"`	Direct command (one-shot)
`claude commit`	Smart commit
`/help`	In-session help
`/clear`	Clear context

Hooks

Hooks trigger scripts automatically before or after Claude Code actions.

Full guide: Claude Code Hooks

MCP (Model Context Protocol)

MCP connects Claude Code to external data sources: databases, APIs, cloud services.

Full guide: Claude Code MCP

Skills and custom commands

Create reusable commands with skills.

Full guide: Skills and commands

Part 3: Power user tips

The ideal workflow

Solid CLAUDE.md - invest time polishing it
Safety hooks - backup before edit, lint after modification
Connected MCP - databases, APIs, deploy tools
Custom skills - automate recurring tasks

Common mistakes

Too vague: “improve my code” → be specific about what to improve
No context: Claude Code reads files, but guide it to the right ones
No review: never merge without understanding what Claude did
Prompts too long: direct instructions > ceremonious mega-prompts

Going further

Install Claude Code - step-by-step installation
CLAUDE.md best practices - configure your project
Hooks - automate your workflow
MCP - connect external services
Skills - create custom commands

This guide is updated regularly. Last update: March 2026.