Everything About Claude: The Complete Beginner-to-Power-User Guide (2026)

Claude can read your files, remember context across chats, build documents and slide decks, browse websites, write and run code, and keep working on a task while you are away. This guide skips the trial-and-error phase. Everything here reflects Claude as of October 2026; features change often, so plan- or region-specific details are flagged.

What Claude is, and which model to pick

Claude is a family of AI models built by Anthropic, an AI safety company. “Claude” is also the name of the app you chat with. The app lets you switch models, and picking the right one is the first skill beginners miss. Bigger models reason better but use more of your plan’s allowance and respond more slowly.

Hard-to-find detail — the Mythos tier. Above Opus sits Anthropic’s newest “Mythos” tier. Claude Mythos 5.1 and Claude Fable 5.1 share the same underlying model; Fable adds extra safety measures for sensitive areas such as biology and cybersecurity, which is why you see Fable in the app. A separate research model, Claude Mythos Preview, is not public and is used by a small group of organisations under Anthropic’s Project Glasswing.

Context window is how much text Claude can hold in mind at once. On current plans it reaches up to 1 million tokens (roughly 750,000 words), varying by model — enough for a long book or a large codebase in one conversation. New to tokens? Read What Are Tokens in LLMs?

Where you can use Claude

One account, many surfaces. Your chats, Projects, skills and connectors follow you across web, desktop and mobile, and all usage draws from one shared pool.

Plans, pricing and how usage limits really work

The Free plan is genuinely useful; Pro at $20/month is the right upgrade for most daily users. Prices are in US dollars, before tax, from Anthropic’s pricing page.

How the limits work (the part nobody explains)

• There is no fixed message count. Usage depends on conversation length, attached files, the model and the features you use.
• Limits reset on a rolling 5-hour window that starts with your first message. Paid plans also have a weekly limit on top.
• Everything draws from one pool — web, desktop, mobile and Claude Code.
• Long chats get expensive. Every new message re-reads the whole conversation, so message 50 of a long thread costs far more than message 1.
• Check your standing any time under Settings > Usage. Hit the limit? Wait for the reset, upgrade, or turn on usage credits (paid plans) at standard API rates.

As a rough guide, Anthropic says a Pro user with short conversations on a lighter model can expect around 45 messages every five hours, often more.

Five ways to stretch your usage

1. Batch your questions — three things in one message, not three messages.
2. Don’t re-upload files in the same chat; Claude already has them.
3. Use Projects for repeated material. Project files are cached, and cached content counts less against your limit.
4. Start fresh chats for new topics. Ask for a short summary, then paste it into a new chat.
5. Match the model to the task. Haiku or Sonnet for simple jobs; save Opus and Fable for hard ones.

The core features, and how to use each one

Claude’s features stack in three layers. Power users set all three up once, then reuse them.

Layer 1: Context — stop re-explaining yourself

• File uploads. PDFs, Word, spreadsheets, images, code, CSVs — including charts and scanned pages. One well-chosen file beats ten loosely related ones.
• Projects. A workspace with its own instructions, reference files and chat history; every chat inside starts with that context. Create one per client, course, product or recurring job. Free: up to 5; Pro and above: unlimited.
• Instructions vs knowledge. Instructions are how Claude behaves (“answer as a reinsurance analyst, use INR, under 200 words”). Knowledge files are what it knows. Keep them separate.
• Memory. On every plan, Claude can remember your role, projects and preferences. View and edit it in Settings, or ask Claude directly to remember or forget something.
• Past-chat search. “What did we decide about the pricing page last week?” pulls back that earlier conversation.
• Incognito chats don’t feed memory or appear in history — ideal for one-off or sensitive questions.
• Preferences and styles. Set tone, format, language and units once in Settings.

Layer 2: Tools — reach the real world

• Web search (all plans) looks up current information and cites sources.
• Research (Pro and up) runs many searches, reads widely and returns a cited report — great for market scans and literature reviews.
• Code execution and file creation. Claude runs Python in a private sandbox to analyse data, build charts and produce real Excel, Word, PowerPoint and PDF files.
• Connectors link Claude to Gmail, Google Drive, Calendar, Slack, Notion, GitHub, Jira, Figma and hundreds more through the open Model Context Protocol (MCP).
• Skills are folders of instructions (sometimes with scripts) that teach Claude one task your way, loaded only when the task calls for them.
• Plugins bundle skills, connectors, commands and helper agents into one install, from a shared directory.
• Browsers and computer use. Claude can fill forms and click through sites in its built-in browser or your Chrome, and operate desktop apps through the desktop app — asking before it sends, buys or submits anything.
• Voice mode for hands-free conversations on mobile and desktop.

Layer 3: Outputs — finished work, not text to copy

• Artifacts — documents, slide decks, spreadsheets, designs, dashboards, whiteboards, task lists, small web apps and calculators, built beside the chat, editable by you and shareable by private link.
• Claude Docs, Slides and Design (all plans) — ask for “a proposal”, “a 10-slide deck” or “a poster” and get a live, editable file that exports to Word, PowerPoint, PDF or images.
• Dashboards (Pro and up, beta) — live KPI pages built from your data.
• Scheduled and long-running tasks (Pro and up) — “Every Monday at 9, summarise my unread client emails” runs on its own, even when you’re offline.

Prompting that actually works

Good prompts read like a brief to a smart new colleague: who it’s for, what done looks like, and what to avoid. Claude follows instructions closely, so vague asks get generic answers. For a deeper dive, see our prompt engineering guide.

The five-part brief

1. Goal — what you want and why.
2. Context — audience, background, constraints.
3. Format — length, structure, tone.
4. Examples — paste one you liked and one you didn’t, and say why.
5. Done criteria — how you’ll judge the result.

Techniques beginners rarely know

• Use XML tags to separate inputs — wrap pasted material in <contract>…</contract> and <instructions>…</instructions>.
• Ask for the output shape — “a table with columns X, Y, Z” or “reply in JSON”.
• Let it think — “think step by step before answering”, or turn on extended thinking for maths, logic and planning.
• Ask Claude to ask you — “Before you start, ask me up to three questions that would change your answer.”
• Give it a role, briefly — “You are a senior reinsurance underwriter reviewing this treaty.”
• Iterate, don’t restart — say what to change rather than rewriting the prompt.
• Ask for a critique — “Review your answer as a sceptical expert and list its weakest points.”
• Turn good prompts into assets — a weekly prompt belongs in Project instructions or a skill.

Hidden gems: the details beginners rarely find

1. Training is opt-out on Free, Pro and Max. Check Settings > Privacy. Team and Enterprise don’t train on your content by default.
2. Connectors are switched on per conversation. Connecting Gmail once doesn’t let every chat see your inbox.
3. Project memory can stay separate, so client work doesn’t mix with personal chats.
4. You can import memory from another assistant instead of rebuilding context.
5. Edit your message instead of arguing. Correcting in a follow-up keeps the bad answer in context and burns usage.
6. Artifacts are private until you share them.
7. “Make it a doc” changes everything — you get something editable and exportable, not text to reformat.
8. Claude keeps working while you’re away on paid plans; check back from your phone.
9. Max plans include free API credits — $100 to $200 a month for building your own tools.
10. Max can’t be billed annually; only Pro and Team get the annual discount.
11. Cancelling deletes nothing. Cancel at least 24 hours before renewal to avoid the next charge.
12. App Store subscriptions are refunded by Apple, not Anthropic. Subscribe on the web to manage billing directly.
13. Attach first, then ask. Claude can’t see a file you only mention.

Workflows: how to leverage Claude in your role

The biggest gains come from repeatable workflows, not one-off questions. Pick the row closest to you and set it up once.

Limitations, privacy and safety

Treat Claude’s output like a capable colleague’s first draft: usually good, always worth checking where it matters.

• What Claude won’t do: serious-harm requests (weapons, malware), reproducing song lyrics or long copyrighted passages, drawing known copyrighted characters, or building fake login pages, receipts or reviews — even “for a demo”.
• It isn’t your lawyer, doctor or financial adviser. It explains options; you decide.
• Protect your data: use incognito for sensitive one-offs, review Settings > Privacy, connect apps only where needed, and read the approval prompts when Claude acts in a browser.
• For company data, prefer Team or Enterprise: no training on your content by default, plus admin controls, SSO and audit logs.

Beginner mistakes, and a 7-day plan to avoid them

Seven common mistakes

1. One endless chat for everything — slow, expensive, confused.
2. Prompts with no context — “write an email” gets a generic email.
3. Never switching models.
4. Copy-pasting instead of asking for a file.
5. Ignoring connectors.
6. Trusting without checking — especially numbers and citations.
7. Rewriting prompts from scratch every week — save them as Project instructions or a skill.

Your first week with Claude

Sources

• Claude plans and pricing — Anthropic (read October 14, 2026)
• About Claude’s Pro plan usage — Claude Help Center
• Connectors, skills and plugins — Claude docs
• Prompt engineering overview — Anthropic docs
• Project Glasswing — Anthropic