Updated May 2026 · Mini-course

Claude AI — Complete Guide

Everything you need to know about Claude — from models and prompting to agents, skills, and commands. Written for beginners, useful for everyone.

What is Claude?

Claude is an AI assistant made by Anthropic, a safety-focused AI company founded in 2021. Claude can read, write, analyze, code, summarize, translate, and reason through problems — much like ChatGPT, but with a different emphasis on safety, nuance, and long-context understanding.

Claude is available on claude.ai (free and paid plans), via the Anthropic API (for developers), and as Claude Code (a CLI tool for software engineers). Each interface gives you the same underlying model but different capabilities.

The name “Claude” comes from Claude Shannon, the mathematician who founded information theory — a nod to Anthropic's research roots.

🌐

claude.ai

Web interface. Free tier available. Start here if you're new.

⚙️

Anthropic API

For developers. Access Claude programmatically in your apps.

💻

Claude Code

CLI tool for software engineers. Runs in your terminal and IDE.

Claude Models Explained

Anthropic releases Claude in three tiers. Think of them as the same brain at different speeds and price points.

🧠

Claude Opus

Most capableSpeed: ★★☆☆☆ · Cost: $$$$

The smartest Claude model. Best for complex reasoning, nuanced writing, multi-step analysis, and tasks where quality matters more than speed.

Best for:

  • Complex research & analysis
  • Long-form writing
  • Advanced coding
  • Strategic planning
  • Multi-step reasoning chains

Claude Sonnet

Best balanceSpeed: ★★★★☆ · Cost: $$$

The everyday Claude. Sonnet is fast, smart, and cost-effective — the right choice for most tasks. It's what most users should default to.

Best for:

  • Everyday writing & editing
  • Code generation
  • Q&A and explanations
  • Summarization
  • Customer support drafts
🚀

Claude Haiku

Fastest & cheapestSpeed: ★★★★★ · Cost: $

Ultra-fast and low-cost. Haiku handles simple tasks at scale — ideal for classifying, tagging, short responses, or high-volume pipelines.

Best for:

  • Classification & tagging
  • Simple Q&A
  • High-volume automation
  • Real-time responses
  • Form validation & routing

Which model should you use?

Start with Sonnet for everything. Switch to Opus when you need deeper reasoning and Sonnet isn't cutting it. Use Haikuwhen you're building an app and need to process thousands of requests cheaply.

How Claude Works — Prompting Basics

Claude works through conversation. You write a message (called a prompt), Claude responds. The quality of your output depends almost entirely on the quality of your input.

1

System prompt (optional, powerful)

A hidden instruction that shapes how Claude behaves for the entire conversation. Used to give Claude a role, set rules, define output format, or restrict topics. This is what AI agents use.

"You are a senior Ruby developer. Always write tests alongside code. Never use deprecated gems."
2

User message

Your actual question or task. The more specific and structured, the better the output.

"Refactor this function to be more readable and add error handling for null inputs."
3

Claude's response

Claude reads both the system prompt and your message, then generates a response that follows all constraints it has been given.

Claude returns refactored code + explanation of changes + unit tests.
4

Context window

Claude remembers everything in the current conversation up to its context limit (200,000 tokens for Claude 3.x — roughly 150,000 words). Beyond that, older messages are forgotten.

You can paste an entire codebase, book, or document and Claude will reason across all of it.

What Are AI Agents?

An AI agent is Claude with a specific identity, role, and ruleset baked in from the start. Instead of asking a generic AI a question, you give Claude a system prompt that makes it behave like a specialist.

Generic Claude vs Agent Claude:

❌ Generic prompt

“Help me grow my startup”

Vague answer. Generic advice. No framework.

✅ With Growth Hacker Agent

“Help me grow my startup”

Analyzes funnel, identifies biggest leverage points, gives a 30-day experiment roadmap with metrics to track.

How to use an agent — 3 steps:

  1. Pick an agent from the library (or write your own system prompt)
  2. Start a new conversation in Claude.ai, ChatGPT, or any AI tool
  3. Paste the system prompt as your first message — then give your task

On this site we have 150+ free agents for marketing, coding, sales, design, research, and more — all ready to copy and paste.

Browse free agents →

What Are Skills?

Skills (in Claude Code — the CLI tool) are reusable slash commands that trigger a pre-defined workflow. Instead of typing the same complex prompt every time, you type /skill-name and Claude executes a complete, structured task.

Example skill invocations:

/review # runs a full code review on the current branch

/seo # audits current page for SEO issues

/init # creates a CLAUDE.md file documenting your codebase

/security-review # checks for security vulnerabilities

Skills vs Agents

Agents define WHO Claude is (persona, rules). Skills define WHAT Claude does (specific task workflow). They work together — a code reviewer agent + a /review skill.

Skills vs Commands

Commands (/help, /clear, /model) control the Claude Code app itself. Skills (/review, /seo) run custom AI-powered workflows you or someone else defined.

Built-in skills

Claude Code ships with built-in skills like /review, /init, /security-review. These are professionally designed workflows ready to use.

Custom skills

You can create your own skills in .claude/commands/ as markdown files. Write the prompt once, use it forever with a single slash command.

Claude Code Commands Reference

These are built-in slash commands that control Claude Code itself — not AI workflows, but app-level controls. Type them directly in the Claude Code terminal.

/help

Show all available commands and how to use them.

/clear

Clear the conversation context and start fresh.

/model

Switch between Claude models (Opus, Sonnet, Haiku) mid-conversation.

/save

Save the current conversation to a file.

/memory

View or edit what Claude remembers about you across sessions.

/settings

Open Claude Code settings to configure behavior, permissions, and hooks.

/review

Run a code review on the current branch or pull request.

/init

Initialize a CLAUDE.md file with documentation for the current codebase.

/mcp

Manage MCP (Model Context Protocol) server connections.

/cost

Show the token usage and cost estimate for the current session.

/compact

Compress the conversation context to save tokens while preserving key info.

/bug

Report a bug or issue with Claude Code directly to Anthropic.

6 Prompting Tips That Actually Work

These techniques work in Claude, ChatGPT, and any other large language model. Master these and you'll get dramatically better output.

🎭

Give Claude a role

Start your prompt with "You are a [role]" — Claude will adopt that persona and respond differently. "You are a senior security engineer reviewing this code for vulnerabilities" gets better output than "check this code."

📐

Be specific about the output format

Tell Claude exactly what you want back: "Give me a bullet list", "Write this as a JSON object", "Format as a markdown table with 3 columns". Vague requests get vague output.

📋

Use examples (few-shot prompting)

Show Claude one or two examples of what you want before asking. "Here is an example of the style I want: [example]. Now write one about X." This is the single most effective prompting technique.

⛓️

Chain tasks step by step

For complex tasks, break them into explicit steps: "First, list the problems. Then, rank them by severity. Then, suggest a fix for the top 3." Claude follows sequential instructions better than open-ended ones.

💭

Ask Claude to think before answering

For reasoning tasks, add "Think step by step before giving your answer." This activates Claude's extended thinking and dramatically improves accuracy on math, logic, and analysis.

🚧

Set constraints

Tell Claude what NOT to do: "Don't use jargon", "Don't exceed 200 words", "Don't suggest paid tools". Constraints shape output as much as instructions do.

Frequently Asked Questions

Is Claude free to use?
Claude has a free tier on claude.ai with usage limits. Claude Pro ($20/month) gives higher limits, priority access, and access to all models. The API is pay-per-use with no monthly fee — you pay only for tokens used. Claude Code (CLI) is separately billed per token.
How is Claude different from ChatGPT?
Both are large language models but with different strengths. Claude tends to be better at nuanced writing, following complex instructions, and handling very long documents (200k token context vs ChatGPT's 128k). ChatGPT has more plugins/integrations and a larger ecosystem. For coding and analysis, they are comparable — try both.
Can Claude access the internet?
By default, Claude cannot browse the web. On claude.ai, Anthropic has added web search as a feature. Via the API, you can give Claude tools (like a search function) that let it fetch live data. Claude Code can run terminal commands and read your local files, but doesn't browse the web unless you configure it with an MCP server that enables web access.
What is MCP (Model Context Protocol)?
MCP is an open standard by Anthropic that lets Claude connect to external tools — databases, APIs, file systems, calendars, and more. With an MCP server configured, Claude can read your Notion, query your database, or call your internal APIs. It's the foundation for truly autonomous AI agents.
Is it safe to paste my code into Claude?
Claude does not use your inputs to train its models (unless you opt in). On claude.ai, Anthropic stores conversations per their privacy policy. For sensitive production code, use the API with zero-data-retention enabled, or use Claude Code locally (which can read files without sending them to the cloud unless you start a conversation).
What is CLAUDE.md?
CLAUDE.md is a file you place in your project root that Claude Code reads automatically at the start of every session. It's where you document your codebase: the tech stack, conventions, what to avoid, how to run tests, etc. Think of it as a permanent brief that Claude always has in context. Run /init to generate one automatically.

Ready to use Claude like a pro?

Pick a free agent, paste it into Claude or ChatGPT, and get specialist-level output in seconds.

Browse 150+ free agents →