Plain-English Walkthrough
Claude Tutorial for Beginners: Your First Hour, Step by Step
Everything you need to go from a blank sign-up page to a saved, useful conversation.
This is a doing-guide, not a features list. It walks through the first hour with Claude in order: creating an account, writing a first prompt, attaching a file, fixing a disappointing answer, and saving your work so it's still there next week. Every step assumes you have never opened Claude before.
To use Claude as a beginner: create a free account at claude.ai, land in the chat window, and type a specific request rather than a vague one. Use the attachment icon in the message box to upload a file (PDF, image, spreadsheet, or document) and ask a question about it directly. If the first answer is off, don't start over, reply in the same conversation with what was wrong and what you actually wanted. To keep useful chats, rename them or move related ones into a Project so the context carries forward.
Published
Key Takeaways
- Claude.ai is a chat window first: you type a request, Claude replies in the same thread, and the whole exchange is preserved as a conversation you can return to.
- The paperclip or attach icon in the message box lets you upload files, including PDFs, images, spreadsheets, Word docs, and ask questions grounded in that specific content.
- A bad first answer is normal and fixable. Reply in the same thread with what was missing rather than closing the chat and starting a new one from scratch.
- Projects let you group related conversations and files under one shared context, so you're not re-explaining your situation in every new chat.
- Artifacts is the side panel Claude uses for substantial output, such as a document, a piece of code, or a chart, that you can view, edit, and revise without it getting buried in chat text.
- Claude is a general-purpose writing and reasoning assistant, not a search engine. It's strongest when you give it a specific task and context, not a bare keyword query.
Which Claude surface should you use for which first-hour task?
| Task | Best surface | Why it fits |
|---|---|---|
| A quick one-off question | A new chat | No setup needed. Type, get an answer, done. Most first prompts belong here. |
| Uploading a document, image, or spreadsheet to ask about | The attach icon in the message box | Claude reads the file's actual content rather than guessing from a description of it. |
| Work that spans several sessions on the same topic | A Project | Shared context and files persist across every chat inside the Project, so you stop re-explaining yourself. |
| A document, piece of code, or chart you'll want to revise repeatedly | Artifacts (opens automatically) | The output lives in a separate panel you can edit directly, instead of scrolling back through chat text to find the latest version. |
| Writing or editing code inside your own project files | Claude Code (a separate tool) | Built for working directly in a codebase with a terminal, not for a first-hour introduction to chat. |
Copy and Paste
Prompts You Can Use Today
A first prompt that actually tests what Claude can do
I'm new to using AI assistants. I want to draft a short email to my landlord asking to extend my lease by three months, staying on the same terms. Keep it polite but direct, under 120 words, and give me two versions: one more formal and one more casual.
Swap the scenario for anything you actually need written. The structure (context, task, constraints, format) is what matters.
Fixing a vague or generic first answer
That's too generic and doesn't mention that I've been a good tenant for four years with no late payments. Rewrite the formal version so it opens by referencing that, and cut it down to five sentences total.
Reply in the same conversation. Do not start a new chat. Claude keeps the earlier context and only needs to know what to change.
Asking a question grounded in an attached file
I've attached my resume as a PDF. Read through it and tell me the three weakest bullet points from a hiring manager's perspective, and for each one, rewrite it to lead with a measurable result instead of a vague responsibility.
Attach the file first using the paperclip icon in the message box, then send this as your message.
Understanding the reasoning behind an answer, not just the answer
Before I use that rewritten paragraph, explain your reasoning: why did you choose to lead with the numbers instead of the job title, and is there a version of this convention that would look strange in a more traditional industry like law or accounting?
This is useful whenever you're not sure if you should trust the output as-is.
Starting a Project to keep related work together
I'm going to use this Project to plan a move to a new city over the next two months. Here's what's true so far: I have a budget of what I've told you, I need to be near public transit, and I have a cat. Keep that context for everything I ask in this Project from now on.
Create the Project first from the sidebar, then open a new chat inside it and send this as the first message.
Asking Claude to critique its own work before you rely on it
Before I send this, review your own draft as if you were a skeptical editor. Point out anything that sounds overconfident, any claim I should double-check myself, and anything that reads as generic AI writing rather than something I'd actually say.
A good habit for anything you're about to send to someone else. Treat the first draft as a draft, not a final answer.
The Hour, Mapped Out
What you'll actually do in your first hour
Five steps, in order. Each one builds on the last.
1. Create an account and open the chat window
Sign up with an email or an existing account, and you land directly in a blank chat with a message box at the bottom.
2. Write a specific first prompt
Not a keyword search. A real request with context, a task, and a format, the way you'd brief a capable colleague.
3. Attach a file and ask about it directly
Use the paperclip icon to upload a PDF, image, or document, then ask a question grounded in what's actually in it.
4. Iterate instead of restarting
When an answer is vague or wrong, reply in the same thread with what to fix. Starting over throws away useful context.
5. Save and organize what's worth keeping
Rename useful chats, or move related ones into a Project so the context and files carry forward automatically.
Before You Start
What Claude actually is, so your first prompt doesn't waste your time
Claude is a conversational assistant, not a search engine. A search engine takes a few keywords and returns links you have to sift through. Claude takes a full instruction, written the way you'd explain a task to a person, and produces a direct answer, draft, or piece of reasoning in response. The difference matters for how you should write your first prompt: a keyword phrase like "resume tips" will get you something generic, while a full sentence describing your actual resume, your actual industry, and what you want changed will get you something usable.
The other thing worth knowing before you start is that Claude doesn't remember you between separate, unrelated conversations by default. Each new chat starts fresh unless you're working inside a Project, which carries shared context forward, or unless you explicitly restate what matters. This isn't a flaw to work around, it's closer to how a good consultant treats client confidentiality between different engagements. You just need to know it going in, so you're not confused when a brand-new chat doesn't recall something you mentioned in a different one yesterday.
Claude also isn't guessing at facts from a fixed script. It reasons through your specific request using the context you give it, which is why the same underlying capability can draft an email, explain a legal term in plain language, debug a spreadsheet formula, or walk through the logic behind a decision. The quality of what comes back tracks closely with the quality and specificity of what you put in.
None of this requires technical knowledge to use well. The skill that matters most for a beginner isn't prompt syntax or special commands. It's the habit of writing requests the way you'd brief someone competent: what you're trying to accomplish, any constraints that matter, and what form you want the answer in.
Step 1
Create your account and get oriented in the chat window
Go to claude.ai and sign up with an email address or an existing account you already use. Once you're signed in, you land directly in a new, empty conversation. There's no dashboard to configure first, no settings you need to touch before you can start. The message box at the bottom of the screen is where everything begins.
Take thirty seconds to notice three things on the screen before you type anything. First, the sidebar on the left holds your conversation history, every chat you have gets saved here automatically, searchable and renameable, so nothing you do in your first hour is thrown away by accident. Second, the message box at the bottom has a paperclip or attach icon next to it, which is how you'll upload files later in this walkthrough. Third, most accounts show a model selector somewhere near the message box, a dropdown that lets you choose which underlying model handles your request. As a beginner, you don't need to make a decision here; the default selection is a sensible one, and you can revisit this choice once you have a feel for how Claude responds.
There's nothing to break by exploring. Every action in the interface, starting a new chat, renaming one, deleting one, is reversible or low-stakes, which makes this a good place to click around for a minute before you write your first real prompt.
Step 2
Write your first prompt (and why it's usually mediocre)
The most common first mistake is typing something short and vague, "write me an email" or "help with my resume", and judging the whole tool by the flat, generic response that comes back. That response isn't a limitation of Claude; it's a direct reflection of how little the prompt gave it to work with. A vague request can only produce a generic answer, because there's no specific situation to respond to.
A better first prompt has four ingredients: who you are or what the situation is, the specific task, any constraints that matter (length, tone, audience, what to avoid), and the format you want back. You don't need to label these explicitly, just write a few sentences that cover them naturally, the way you would if you were asking a sharp colleague for a favor rather than typing a search query.
Send it, read the response fully, and resist the urge to judge it against perfection on the first try. Claude's replies are drafts meant to be worked with, not final answers meant to be accepted or rejected outright, which is exactly what the next step covers.
Step 3
Attach a file and ask a real question about it
This is where Claude stops feeling like a chatbot and starts feeling like a tool you can actually put to work.
Find the attach icon
It sits directly next to the message box, usually shown as a paperclip. Click it to open a file picker.
Pick something with real stakes
A resume, a lease, a spreadsheet of expenses, a screenshot of an error message, anything you'd genuinely like feedback or analysis on.
Ask a specific question about it
Not "what do you think" but a targeted question: what's weakest, what's missing, what would a specific kind of reader flag.
Check the answer against the file
Scroll back to the actual document and verify Claude referenced real content from it, not a generic guess based on the file type.
Step 4
When the answer is wrong or vague, iterate, don't restart
This is the single habit that separates people who find Claude genuinely useful from people who try it once, get an underwhelming answer, and give up. A disappointing first response is not a dead end. It's a starting point for a second, much better prompt, sent in the same conversation.
The reason to stay in the same thread rather than opening a new chat is that Claude retains everything said earlier in that conversation. If you close it and start over, you lose that context and have to re-explain your situation from scratch. If you reply in place, you can simply say what was wrong, too long, too generic, missed a key detail, wrong tone, and Claude revises with full knowledge of what it already tried.
Treat the first response the way a writer treats a first draft: expected to need revision, not expected to be final. Two or three rounds of "here's what to fix" inside one conversation will almost always outperform judging the tool on a single, unrefined attempt.
This same iteration habit applies to Artifacts. If your prompt produces a substantial piece of output, a formatted document, a block of code, a chart, Claude will often open it in a separate panel next to the chat rather than inline in the conversation. You can ask for changes to that artifact the same way: describe what to fix, and the panel updates in place rather than generating a duplicate copy buried further down the chat.
Step 5
Save your work so it isn't a dead-end conversation
Every conversation is saved automatically to the sidebar, but "automatically saved" and "actually findable in three weeks" are different things. Two habits fix this. First, rename any conversation worth keeping, click into its options and give it a name you'd actually recognize later, rather than leaving it titled by whatever the first message happened to be. Second, if you find yourself returning to the same topic across multiple separate chats, planning something over weeks, working through a recurring type of task, create a Project instead of letting the work scatter across unconnected conversations.
A Project is a container with its own shared context and its own file storage. Anything you upload or establish as true inside a Project, your constraints, your preferences, background facts about your situation, is available to every conversation you start inside it, without you needing to repeat yourself each time. For a beginner, the test for whether something deserves a Project is simple: if you can imagine asking a related but different question about the same situation next week, it belongs in a Project rather than a one-off chat.
This is also the point where it's worth glancing at your conversation history and noticing a pattern in what you asked for in your first hour. That pattern is a reasonable signal for what to build a Project around going forward, most people's first hour reveals one or two recurring needs (work writing, a personal project, learning a skill) worth giving a permanent home.
Honest Limits
What Claude won't do well in your first hour, and why that's fine
Claude doesn't have live access to today's news, stock prices, or anything happening after its training cutoff unless it's explicitly using a search capability for that request, and even then it's worth treating time-sensitive claims as things to verify yourself before acting on them. If your first prompt is about something that changed this week, don't be surprised if the answer needs a follow-up asking Claude to search for current information, or needs independent verification on your part.
It also won't reliably know specific, private facts about you that you haven't told it, your bank balance, your calendar, your company's internal policies, unless you've provided that information in the conversation or connected a relevant integration. A beginner mistake is asking a question that assumes Claude has context it was never given, then concluding the tool is unreliable when the real issue is a missing piece of information in the prompt.
And for anything with real consequences, a legal document, a medical question, a number that will end up in a financial decision, the right posture is to use Claude to draft, explain, and accelerate your own thinking, and then verify the specifics yourself or with a qualified professional before acting. That's not a weakness particular to Claude; it's the same posture worth taking with any single source, human or AI, on a high-stakes question.
After Your First Hour
What's worth trying next
Once the basics feel natural, these are the next things people usually reach for.
Set up a Project for your main use case
Whatever came up most in your first hour, work writing, a personal project, studying a topic, give it a permanent home with shared context.
Try a multi-step task instead of a single question
Ask Claude to plan something with several parts, then work through it piece by piece in one conversation rather than one prompt at a time.
Look into Claude Code if you write software
A separate, terminal-based tool built for working directly inside a codebase, a different surface from the chat window covered here.
Practice giving worse prompts on purpose
Try one vague prompt and one specific prompt on the same task, side by side, to build an intuition for what makes the difference.
Reading a tutorial and being fluent with Claude are different things
You now know the five steps, but the habits that make Claude genuinely useful, writing specific prompts, iterating instead of restarting, knowing when to reach for a Project, only stick with practice. The free Nightschool AI curriculum is hands-on from the first lesson, built around real tasks instead of feature tours.
Frequently Asked Questions
Do I need to pay to try Claude?
You can create a free account at claude.ai and use it without payment. Paid plans exist for people who want higher usage limits or additional capabilities, but everything covered in this first-hour walkthrough, chatting, attaching files, iterating on answers, using Projects, is accessible on a free account.
What's the difference between Claude and ChatGPT for a beginner?
Both are conversational AI assistants with a similar core interaction: you type a request, you get a response, you can attach files and iterate in the same conversation. The meaningful differences for a beginner show up less in the first hour and more over time, in writing style, how each handles long documents, and how each product organizes ongoing work (Claude's Projects versus the other's equivalent).
Why did my first Claude answer feel generic or unhelpful?
Almost always because the prompt itself was generic. A short, vague request like 'help me write an email' gives Claude nothing specific to respond to, so it returns something equally generic. Add real context, who it's to, what you're trying to accomplish, any constraints, and the answer sharpens immediately.
Can Claude read a PDF, image, or spreadsheet I upload?
Yes. Click the attach icon (usually a paperclip) next to the message box, select your file, and ask a specific question about it. Claude reads the actual content of the file rather than guessing based on the file type, so questions grounded in what's really in the document tend to get much more useful answers.
Does Claude remember previous conversations?
Not by default across separate, unrelated chats, each new conversation generally starts fresh. Inside a Project, however, shared context and uploaded files carry forward automatically to every conversation started within that Project, which is the main reason to use one for ongoing or recurring work.
What should my very first prompt be?
Something with real stakes to you, not a generic test question. Pick an actual task, a real email you need to write, a real document you want feedback on, and describe the situation, the task, any constraints, and the format you want back in a few full sentences rather than a short phrase.
What is a Project in Claude, and do I need one right away?
A Project is a container that groups related conversations and files under shared context, so you don't have to re-explain your situation in every new chat. You don't need one for your first hour, start with regular chats, and create a Project once you notice you're returning to the same topic repeatedly.
What is Artifacts and when does it show up?
Artifacts is a separate panel that opens next to the chat when Claude produces substantial output, a formatted document, a piece of code, a chart. Instead of that content sitting inline in the chat text, it lives in an editable panel you can revise directly by describing what to change.
How do I get Claude to fix a bad answer without starting over?
Reply in the same conversation rather than opening a new chat. Say specifically what was wrong, too vague, too long, missed a detail, wrong tone, and Claude revises with full knowledge of what it already produced, which almost always works better than restarting from a blank chat.
Can I trust Claude's answers without checking them?
Treat Claude as a strong first draft and reasoning partner, not a final authority, especially on anything time-sensitive, highly specific, or high-stakes (legal, medical, financial). For everyday writing and thinking tasks it's reliable as a starting point; for consequential decisions, verify specifics yourself or with a qualified professional.
Is there a mobile app, or do I need to use a browser?
Claude is available both as a mobile app and through a browser at claude.ai, and conversations sync between them since they're tied to the same account. For a first-hour walkthrough, either works identically, the steps in this guide apply the same way on both.
What's the single most useful habit for a Claude beginner to build?
Writing specific prompts and iterating in place. Almost every disappointing first experience with Claude traces back to one of these two things: a vague first prompt, or abandoning a conversation instead of correcting it. Fix those two habits and the tool becomes noticeably more useful within the same first hour.
You've done the first hour. Now build the habit.
The gap between knowing the steps and using them well closes with practice, not more reading.