GuidesHow-to
How to Use Claude: A Complete Beginner's Guide (2026)
Everything you need to start with Claude — the free tier, the five-hour limit nobody explains, Projects, Artifacts, and the one thing it deliberately cannot do.
Brand marks are the property of their respective owners
Claude is very good at a narrow set of things and slightly worse than its reputation at others. This guide covers what it actually does, how to get started free, and the three things that confuse almost every new user — including the usage limit that makes people think it is broken.
What Claude is, in plain English
Claude is an AI assistant made by Anthropic. You type; it answers. In that sense it is the same shape as ChatGPT or Gemini.
Where it differs is temperament. Claude is noticeably better at long documents, careful editing and prose that sounds like a person rather than a press release. It is the tool writers tend to keep after trying several. Our full assessment is in the Claude review, and we compare it head to head in ChatGPT vs Claude for writing.
Getting started, free
- Go to claude.ai and sign up. No card is required for the free plan.
- Type a question in the box. That is genuinely the whole onboarding.
- Upload a file by dragging it into the chat — PDFs, documents, spreadsheets, images.
The free plan gives you the current models, file uploads, web search and Artifacts. What it gives you less of is usage, which brings us to the thing nobody explains.
The five-hour window — read this before you get annoyed
The single most common “Claude is broken” complaint is not a bug.
Anthropic meters usage on a rolling five-hour session window. Paid plans add weekly limits on top of that. When you hit either, Claude stops and tells you when it resets.
Three consequences worth knowing on day one:
- It resets by itself. You do not need to do anything, and nothing is wrong with your account.
- Everything shares one pool. Your activity on the web app, the desktop app, mobile and Claude Code all draw from the same budget. A heavy coding session in the morning is usage your writing does not get in the afternoon — which surprises people who assume the coding tool is billed separately.
- Long conversations cost more than short ones. Each message carries the whole conversation back to the model, so a thread you have been adding to all day is expensive per message. Starting a fresh chat for a new topic is not just tidier, it is cheaper.
That last point is the practical tip most guides skip. If you are burning through your limit faster than you expect, it is usually one enormous conversation rather than volume.
The two features that change how it feels
Most beginners use Claude as a chat box and never find these. Both are worth five minutes.
Artifacts
When Claude produces something substantial — a document, a table, a chunk of code — it opens in a side panel instead of scrolling past in the chat. You can edit it there, ask for changes, and keep the conversation separate from the thing you are making.
The moment this clicks is the moment Claude stops feeling like a chatbot and starts feeling like a workspace.
Projects
A Project is a container for a body of work: its own files, its own standing instructions, its own conversations. Upload the style guide, the brief and the background documents once, and every conversation inside that Project already knows them.
If you find yourself re-explaining context at the start of every chat, you want a Project. This is the highest-value five minutes a new user can spend.
What Claude deliberately does not do
It does not generate images. Claude reads images you give it and cannot make new ones. People arriving from ChatGPT expect this and are surprised.
If you need images, pair Claude with a dedicated tool — our best AI image tools ranking covers the realistic options, and Midjourney vs FLUX settles the most common choice.
It is cautious. Claude refuses more edge cases than its rivals, and occasionally refuses something reasonable. That is a deliberate design position rather than a bug, and it is the trade that comes with the careful tone people like it for.
Getting better answers
The prompt matters far more than the model. Four habits do most of the work:
Say who it is for and what it is for. “Rewrite this for a customer who is already annoyed” beats “make this better” every time.
Give it the real material. Paste the actual document. Claude handles long inputs unusually well, and a real example is worth a paragraph of description.
Ask for one thing at a time. “Write the whole report” produces generic output. “Write section two only, in my voice, short sentences” produces something you can use.
Ask it to criticise its own work. “What have I claimed here that I cannot back up?” and “what is the weakest paragraph?” are the two most useful follow-ups available, and almost nobody uses them.
Free or paid?
| Free | Pro | Max | |
|---|---|---|---|
| Price | $0 | $20/mo, or $17 billed annually | From $100/mo, and $200/mo |
| Usage | Enough for a few sessions a day | Steady daily use | 5x or 20x Pro |
| Claude Code | Not included | Included | Included |
| Best for | Trying it, occasional work | Most people who use it daily | Heavy agentic work |
Checked against Anthropic’s pricing page on 2 August 2026.
The honest advice: stay free until it annoys you. The free tier is genuinely capable, and the signal that you need Pro is hitting the five-hour limit repeatedly during work you were going to do anyway. Upgrading before that is buying capacity you have not needed yet.
One caveat if you write code: Claude Code is not in the free plan, and there is no trial route. That is covered in Cline vs Claude Code and GitHub Copilot vs Claude Code.
Five things to try today
- Paste a long document and ask: “Summarise the argument, then list what it assumes without proving.”
- Paste something you wrote and ask: “Edit for clarity but keep my voice. Show me what you changed and why.”
- Start a Project for something ongoing, upload the background once, and see how different the second conversation feels.
- Ask it to argue against a decision you have already made.
- Give it a messy set of notes and ask for an outline only — no prose. Fix the outline yourself, then have it write section by section.
The one rule
Claude states wrong things with complete confidence, exactly like every other model. It is more careful than most, which makes the mistakes it does make harder to spot rather than rarer.
So: check anything you would be embarrassed to be wrong about. Names, dates, numbers, quotes and anything legal or medical. If you need answers with sources attached, Perplexity is built for that job and Claude is not — see how to use Perplexity for research.
Where to go next
- Claude review — our full assessment, pros, cons and pricing
- ChatGPT vs Claude for writing — the comparison most people actually want
- Best AI chatbots — where Claude sits against the field
- What you actually get on free AI plans — the free tiers, honestly compared