How to use Claude: your first hour, step by step

What "using Claude" looks like

You go to claude.ai, sign up with an email, Google, or Microsoft account, and type a question or task into a chat box, the way you'd text a coworker. No credit card, no trial clock. Claude answers, you push back or ask for changes, and the conversation keeps going until you have something usable. That's it. There's no command syntax to memorize and no plugin to install first. The skill that actually matters is being specific about what you want, the same way you'd be specific with a new hire on their first day.

This guide walks through a real first session: signing up, sending your first prompt, and trying a handful of tasks that show you what Claude can and can't do.

Your first session, step by step

That's a full first session. Nothing here requires reading documentation first.

Three things to try today

Each of these exercises a different part of what Claude actually does, and each one you can verify yourself without any AI expertise.

Summarize a document you already have. Click the "+" button, attach a PDF, DOCX, or CSV, and ask for a summary or "pull out the three most important numbers." You'll know instantly if it's right because you can check it against the original. Claude supports files up to 500MB each, up to 20 per chat.

Draft an email you've been avoiding. Describe the situation and the tone you want, then ask for two or three versions instead of one. Comparing options teaches you more about how Claude responds to instructions than a single output does. This is also where the biggest beginner mistake shows up, covered below.

Ask a question about something happening right now, with web search on. Toggle the slider/"+" menu to turn on Web search, then ask something time-sensitive: a current price, a recent announcement, "what's the latest on X." Watch for the "Searching the web…" indicator and the source links Claude cites underneath its answer. That's the moment web search stops being an abstract feature and becomes something you've actually seen work.

If any of those three go well, a couple more are worth a look once you're comfortable. Fix or explain a spreadsheet formula: paste the formula or upload the sheet and describe what's wrong, and you'll see Claude reason over the actual data instead of just describing prose back to you. Start a Project for something recurring, like a job or a course, where you upload a few reference files once instead of re-explaining context every chat. And if typing isn't your preference, voice mode is worth a try too: tap the sound-wave icon in the lower-right of the chat window (or in the mobile app's text field) and talk instead of type. It's a beta feature available on every plan, on web and mobile.

How to write a prompt that actually works

The single biggest lever a beginner has isn't a setting, it's how the request is phrased. Anthropic's own prompting guidance for developers makes a point that applies just as much in the regular chat window: briefing Claude like a new hire, not typing into a search bar, changes the output. Even one sentence of role or context shifts the answer.

Vague in, vague out. "Help me with this email" tells Claude nothing about what a good version looks like. "Rewrite this email to be more direct, keep it under 100 words, and drop the apology in the first line" gives Claude something to aim at. The gap between those two prompts is the gap between an answer you have to rewrite yourself and one you can send as-is.

Give concrete, specific context instead of a general instruction. "Organize today's tasks" leaves too much to guess. "List today's unfinished tasks, sorted by start time, then priority, then how long each one will take" tells Claude exactly what "organized" means to you, and you'll get an output that's actually usable instead of one more round of back-and-forth.

One task per message. Asking for a summary, a reply draft, and a title suggestion all in one prompt tends to produce a mediocre version of all three. Ask for one, get a good one, then ask for the next.

Treat the first answer as a draft, not a verdict. If it's close but not right, say what's wrong and ask for a revision. That's faster than starting over, and it's how the conversation is meant to work.

Say what to avoid, not just what you want. "No corporate jargon" or "skip the bullet points" is often more useful than another paragraph describing the tone you're after.

Start a new chat once a conversation gets long. Somewhere past 15 to 20 messages in one thread, earlier decisions can get crowded out by everything that came after. For a one-off task, start fresh. For something recurring, that's what a Project is for.

“Claude answers the request you actually made, not what you typed.”

Where beginners get stuck

Claude isn't going to guess what you didn't say, and it has real edges worth knowing about before you hit them.

• It won't infer unstated context. If you ask it to "fix this," it doesn't know what "fixed" means to you unless you say what's wrong.
• Free-tier usage is capped, and Anthropic doesn't publish the exact number. You'll hit a usage limit that resets on a rolling window rather than at midnight, and the cap depends on message length, attachments, and how busy the system is. If you want the fuller breakdown, see is Claude actually free.
• It can be confidently wrong. On anything you can't personally verify (a legal question, a niche technical detail), treat the answer as a first draft to check, not a citation.
• Long, meandering conversations lose the thread. This isn't a bug so much as a reason to start a new chat rather than one giant one.
• It's not the only chatbot, and it's not automatically the right one for every task. If you're also weighing ChatGPT, Claude vs. ChatGPT walks through where each one actually wins.

None of these are dealbreakers. They're just the shape of the tool, and knowing them up front saves you a frustrating first week.

Tested on Claude Sonnet 5, July 2026.

A quick note on who's telling you this

This site isn't affiliated with Anthropic, the company that makes Claude. We're an independent guide, and nobody's paying us to make any particular feature sound better than it is. When something is a genuine limitation, like the unpublished free-tier cap, we say so instead of glossing over it.

Is Claude free to use?

Yes. Claude.ai has a Free plan that costs $0, with no credit card required and no trial period that expires. You get chat, file uploads, web search, and up to five Projects. Usage is capped, but the cap resets on a rolling basis rather than cutting you off for good.

Which Claude model do I get for free?

Sonnet 5, the same model that's the default for Pro subscribers too. According to Anthropic's own announcement, Sonnet 5 "is available across all plans: it is the default model for Free and Pro plans." Free accounts don't get Opus-tier reasoning, which requires a Pro or Max subscription.

What can I do with Claude on the first try?

Summarize a document, draft an email in a specific tone, explain a concept in plain language, fix a spreadsheet formula, or ask a current-events question with web search toggled on. All five work on the free plan with zero setup beyond signing up.

Do I need to learn special prompt syntax to use Claude?

No. Claude reads plain English. The skill worth learning is being specific: state the task, the context, and what a good result looks like, the same way you'd brief a new coworker rather than type keywords into a search engine.

For the full path from here, getting started has the broader setup walkthrough, and at work covers using Claude day to day on the job once your first session is behind you. And if you're wondering whether an official credential exists, Claude certification covers what's real.