Learn Claude.
The most thoughtful AI assistant. Learn how to use Claude for deep writing, careful analysis, long documents, and coding with precision.
Open Claude ↗- 01Go to claude.ai and click "Sign up" with your email or Google account.
- 02Verify your email and complete the short onboarding flow.
- 03Start chatting on the free tier with Claude Sonnet, or upgrade to Pro for extended thinking and higher limits.
- 04Download the Claude Desktop app for Mac or Windows for quick access, or use claude.ai in your browser.
- 05Try Projects to organize conversations with persistent context, or use Artifacts for code and document previews.
| Plan | Price | What you get |
|---|---|---|
| Free | $0 | Claude Sonnet 4.6 access · Limited daily messages · File uploads (PDFs, images, code) · Basic web search |
| Pro | $20/mo | Claude Opus 4.8 access · Extended thinking mode · Higher usage limits · Projects with persistent context · Claude Code (Sonnet) |
| Max | $100-200/mo | Everything in Pro · 5x-20x higher limits · Persistent memory · Claude Code (Opus) · Early feature access |
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."
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>
※ 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]
※ 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
※ 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.
※ 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.
※ Specifying the exact stack (Next.js 14, App Router, NextAuth v5, TypeScript) prevents Claude from giving outdated or generic answers.