Claude tutorial: setup, strengths, costs, mistakes, prompts

Create the account

claude.ai on the web, or the desktop, iOS or Android app. Free is $0 (Anthropic).

Settings worth changing

Settings > Profile: preferences for every chat (who you are, answer style). Settings > Privacy: the model-training choice (listed as opt-out on Free/Pro/Max). Settings > Usage: where you see your limits.

Projects

Free allows up to 5 projects. Put reference files and project instructions in one, so every chat in it starts with that knowledge.

Connectors

Connectors are listed on every plan. Add only the apps a task needs, and test with a read-only request first.

Step 1: Sign in

Claude works on the web, desktop, iOS and Android, and the Free plan costs $0 (Anthropic).

Step 2: Start a chat

Ask a plain question. Free includes web search, memory, file creation and code execution.

Step 3: Use projects

Free allows up to 5 projects, and paid plans have more.

Writing

Say who it's for, the length, and the tone, e.g. "Write a 100-word thank-you note to a customer, warm, no jargon." Then ask Claude to shorten or rewrite it.

Research

Ask Claude for sources, then open them. Don't rely on a summary you haven't checked.

Files

Upload only what you're OK sharing, and tell Claude exactly what to pull out, e.g. "list every date and dollar amount in this PDF."

Good at

• Long careful writing and editing that follows detailed instructions
• Reading long documents and pulling specifics
• Code, including Claude Code on paid plans
• Creating files (docs, spreadsheets) and running code

Bad at / careful with

• Hitting usage limits mid-task: limits reset on a rolling 5-hour window, plus weekly limits on paid plans
• Facts after its training without search turned on
• Being agreeable: it may accept your wrong premise - ask it to argue against you

How limits work

Rolling 5-hour windows plus weekly limits on paid plans; Claude Code shares the same limits as chat (Anthropic). Paid plans can turn on usage credits at API rates.

Cheaper habits

• Long files cost more every message they stay in the chat: start a new chat once you've extracted what you need
• Use Research only when needed
• Watch Settings > Usage before a big job

Prices and refunds

Pro $20/month or $17/month billed annually; Max from $100/month. Payments "generally non-refundable"; cancel 24h before renewal. claude.com/pricing

Burning the 5-hour window on one giant chat

Fix: split jobs, start fresh chats.

Leading questions

"Isn't X true?" invites agreement. Fix: "Give the strongest case for and against X."

Turning on extra usage without a cap

Fix: decide a monthly amount before enabling usage credits.

Document review

Bad: "check this contract"

Good: "I'm the tenant. List every clause in this lease that costs me money or limits me, quote it, give the section number, and rate risk low/med/high. Not legal advice; flag what to ask a lawyer."

Why: role, criteria, quotes and section numbers make it checkable.

Push back

"Here's my plan. Find the 3 biggest weaknesses. Don't be polite."

Connectors and skills

Connectors, Skills and Memory are listed on every plan, including Free.

Claude Code

Claude Code is included on paid plans and shares the same usage limits as chat (Anthropic FAQ).

Research

Research is a paid-plan feature, available on Pro and Max.

Voice

Voice mode is listed on every plan.

Plans

• Free: $0
• Pro: $20/month billed monthly, or $17/month billed annually ($200 up front)
• Max: from $100/month (5x or 20x Pro usage)
• Team: for 2 to 150 people
• Enterprise: $20 per seat per month, plus usage billed at API rates

Prices exclude tax. Source: claude.com/pricing.

Usage limits

Limits reset on a rolling five-hour window, and paid plans add weekly limits. You can check yours in Settings > Usage. On paid plans you can turn on usage credits at API rates.

Refunds

Anthropic says payments are "generally non-refundable" except where its terms or local law require. Cancel at least 24 hours before renewal to avoid the next charge.

Training

Anthropic's comparison table lists model training as "Opt-out" on Free, Pro and Max.