Learn Claude

Master Claude

The most thoughtful AI assistant. Learn how to use Claude for deep writing, careful analysis, long documents, and coding with precision.

Free tier available; Pro at $20/moLong-form writing, deep analysis, research, coding with large codebases

Getting Started

Go to claude.ai and click "Sign up" with your email or Google account.

Verify your email and complete the short onboarding flow.

Start chatting on the free tier with Claude Sonnet, or upgrade to Pro for extended thinking and higher limits.

Download the Claude Desktop app for Mac or Windows for quick access, or use claude.ai in your browser.

Try Projects to organize conversations with persistent context, or use Artifacts for code and document previews.

Key Features

200K Token Context Window

Claude can process entire books, lengthy codebases, and massive documents in a single conversation. Upload a 500-page PDF and ask questions about any part of it. No other major model matches this for long-document work.

Extended Thinking

For complex problems, Claude can "think out loud" before responding. This produces more accurate answers for math, logic, coding challenges, and multi-step reasoning. You can see its thought process unfold.

Projects

Organize your work into Projects with persistent instructions and uploaded reference files. Every conversation in a Project shares the same context, so Claude always knows your preferences, style guide, or codebase.

Artifacts

Claude can create interactive previews of code, documents, and diagrams right in the chat. See a live preview of HTML/CSS, a formatted markdown document, or a React component without leaving the conversation.

Careful, Nuanced Responses

Claude is designed to be thoughtful and honest. It acknowledges uncertainty, presents multiple perspectives, and avoids making things up. For tasks where accuracy matters more than speed, this is a major advantage.

Claude Desktop App

A native desktop app for Mac and Windows that puts Claude right on your computer. Access Claude without opening a browser, use keyboard shortcuts to summon it instantly, and connect it to local tools and files through MCP (Model Context Protocol) integrations.

Claude Code (Developer Tool)

A command-line tool that lets Claude work directly in your codebase. It can read files, run commands, make edits, and handle complex multi-file changes. Available for developers who want AI-powered coding at the terminal level.

Prompting Tips for Claude

Use XML Tags for Structure

Claude responds especially well to XML-tagged prompts. Wrap different parts of your input in descriptive tags to help Claude understand what each section is for.

<context>
I'm writing a quarterly report for our investors.
</context>

<task>
Summarize these three data points into a professional paragraph.
</task>

<data>
Revenue: $4.2M (+18% YoY)
Churn: 3.1% (down from 4.7%)
New customers: 847
</data>

Set Up Projects for Recurring Work

Instead of repeating context every conversation, create a Project. Upload your style guide, codebase, or reference docs once. Every new conversation in that Project will have full context automatically.

Ask for Step-by-Step Reasoning

Claude produces better results when you ask it to reason through problems step by step. For complex tasks, explicitly request a thinking process before the final answer.

Instead of: "What's the best database for my app?"
Try: "I'm building a real-time chat app with 10K concurrent users. Walk me through the trade-offs between PostgreSQL, MongoDB, and Redis for this use case. Then give your recommendation with reasoning."

Use Extended Thinking for Hard Problems

When a task involves complex logic, math, or multi-step analysis, enable extended thinking (Pro feature). Claude will reason through the problem internally before giving you a more accurate answer. This is especially useful for code debugging and data analysis.

Give Claude a Role

Claude adapts its communication style based on role instructions. Tell it who it should be, and it will adjust depth, terminology, and tone accordingly.

"You are a senior backend engineer reviewing my code. Be direct, point out potential bugs, and suggest improvements. Don't explain basic concepts I already know."

Example Prompts

Copy these prompts to try in Claude. Each one is designed to showcase a specific strength.

Writing

<context> I'm a product manager writing release notes for our B2B SaaS customers. </context> <task> Turn these technical changes into user-friendly release notes. Use our standard format: headline, one-paragraph summary, bullet list of changes, and a "What this means for you" section. </task> <changes> - Migrated auth to OAuth 2.1 - Added batch export API endpoint - Fixed N+1 query in dashboard loading - Reduced p95 latency from 800ms to 200ms </changes>

Why this works: XML structure helps Claude distinguish context, task, and data. The format specification ensures consistent output you can ship directly.

Code Review

Review this React component for bugs, performance issues, and best practice violations. Focus on: unnecessary re-renders, missing error handling, and accessibility. Rate severity as critical/warning/suggestion. [paste component code]

Why this works: Specific review criteria prevent generic feedback. The severity rating makes results actionable for prioritization.

Analysis

I'm uploading our company's Q3 financial report (PDF). Please: 1. Summarize the key metrics in a table 2. Identify the 3 most significant changes from Q2 3. Flag anything that looks concerning 4. Write a 2-paragraph executive summary suitable for a board meeting

Why this works: Leverages the massive context window for full document analysis. Numbered tasks give clear structure to the output.

Research

Compare the three leading approaches to AI safety alignment: RLHF, constitutional AI, and debate. For each, explain the core mechanism, give one real-world example, list strengths and weaknesses, and rate the current maturity level. Present as a comparison table.

Why this works: Claude excels at balanced, thorough comparisons. Requesting a table keeps the output scannable despite the depth.

Coding

I have a Next.js 14 app using the App Router. I need to add authentication with NextAuth.js v5. Walk me through the setup step by step: 1. What packages to install 2. The auth configuration file 3. A middleware for protected routes 4. A sign-in page component Use TypeScript. Show complete, runnable code for each file.

Why this works: Specifying the exact stack (Next.js 14, App Router, NextAuth v5, TypeScript) prevents Claude from giving outdated or generic answers.