Plain-English Explainer
How to Use Claude AI: A Complete Practical Walkthrough
Everything you need to open a new chat and actually get something useful out of it.
This page walks through Claude.ai from the first login to the features people skip past without realizing what they do. You'll learn how the chat interface works, how to write prompts that get you better answers on the first try, how Projects and file uploads change what Claude can help with, and where Claude tends to fall short so you're not caught off guard.
To use Claude AI, go to claude.ai, create a free account, and start typing in the message box, no special syntax required. Claude responds in plain conversation, can read files you upload or paste in, remembers the current conversation as you go, and can build documents, code, or other outputs directly in a side panel called Artifacts. For recurring work, Projects let you group chats with shared files and instructions so you don't repeat context every time.
Published
Key Takeaways
- Claude.ai is the main entry point; you sign up free and start chatting immediately, no technical setup required.
- Prompts work best when they specify the role, the context, the audience, and the output format you want back.
- Projects group related chats under shared files and instructions so you stop re-explaining the same background.
- Artifacts is the side panel where Claude builds documents, code, and other standalone outputs you can edit and export.
- Claude can read files (PDFs, spreadsheets, images) you upload directly into a chat, not just answer from memory.
- Claude is strong at writing, analysis, and code; it can still get specific facts wrong, so verify anything that matters.
Claude.ai Features at a Glance
| Feature | What it does | When to use it |
|---|---|---|
| Chat | A single conversation thread with Claude, remembering everything said earlier in that thread | Quick questions, one-off writing, brainstorming, back-and-forth analysis |
| Projects | A container for multiple chats that share uploaded files and a set of custom instructions | Recurring work like a course you're writing, a codebase you're maintaining, or a role you repeat weekly |
| Artifacts | A side panel where Claude renders standalone content, documents, code, simple web pages, diagrams, separate from the chat log | Any output you want to keep editing, copy out cleanly, or view rendered rather than as raw text |
| File uploads | Attach PDFs, spreadsheets, images, or text files directly into a chat for Claude to read and reference | Analyzing a document, extracting data from a spreadsheet, or asking about an image |
| Claude Code | A separate, terminal-based tool for working inside real codebases: reading files, running commands, making edits | Software development tasks that go beyond copy-pasting code snippets into chat |
Copy and Paste
Prompts You Can Use Today
Get a structured first draft instead of a vague one
Act as an editor who specializes in concise business writing. I'm going to paste a rough draft of an email to a client explaining a project delay. Rewrite it in under 150 words, keep an apologetic but confident tone, and end with a clear next step and date. Here's the draft:
Swap the role, tone, and length constraint for your own writing task.
Ask Claude to think through a decision instead of just answering
I'm deciding between two job offers and want to think this through carefully rather than just get a gut reaction. Ask me clarifying questions one at a time about compensation, growth, and lifestyle fit, then once you have enough information, lay out the tradeoffs in a short table.
Useful for any decision with multiple factors, replace the topic and criteria.
Have Claude analyze an uploaded file instead of guessing
I've attached a CSV of last quarter's sales by region. Identify the three regions with the sharpest month-over-month decline, explain what in the data suggests why, and flag any rows that look like data entry errors rather than real trends.
Upload the file to the chat first, then send this prompt referencing it.
Get working code with an explanation, not just a code block
Write a Python function that takes a list of dictionaries representing invoices, each with a 'due_date' and 'amount', and returns the total amount overdue as of today. Include type hints, handle malformed date strings without crashing, and explain your approach in two sentences before the code.
Replace the language, data shape, and edge cases with your actual problem.
Set up a Project for a recurring role
You are helping me manage weekly status updates for a five-person marketing team. From now on, when I paste raw notes from our standup, reformat them into three sections: Shipped This Week, Blocked, and Next Week's Priorities, using bullet points and no more than one line per item.
Paste this as the custom instructions field when creating a Project, then reuse the project for every week's update.
Use Artifacts to get an editable document, not just chat text
Create a one-page project brief as a formatted document for a client-facing website redesign. Include sections for Goals, Timeline, Deliverables, and Out of Scope, and use placeholder text I can replace with real details.
Claude will typically open this in the Artifacts panel where you can keep editing it directly.
Getting Started
What you actually see when you open Claude.ai
The interface is intentionally sparse. Here's what each piece is for.
The message box
A plain text field at the bottom of the screen. Type in full sentences the way you'd explain something to a colleague, there's no command syntax to memorize.
The attachment icon
Lets you upload files directly into the conversation. Claude reads the contents and can reference specific parts of what you attached in its answer.
The sidebar
Lists your past chats and any Projects you've created, so you can pick up an earlier conversation exactly where you left it.
The model selector
A dropdown to choose which Claude model handles the conversation, faster models for quick tasks, more capable ones for harder reasoning or longer documents.
Artifacts panel
Opens automatically on the right side of the screen when Claude produces something meant to stand alone, like a document or a piece of code, rather than just conversational text.
The Basics
Writing a prompt that actually works
The single biggest difference between a mediocre answer and a genuinely useful one is how much context you give Claude up front. A prompt like "write me a marketing email" forces Claude to guess your audience, your tone, your product, and your length constraints, and it will guess, rather than ask, unless you tell it to check with you first. A prompt like "write a 100-word marketing email to existing customers announcing a price increase, keeping the tone apologetic but not defensive, and end with a specific date the change takes effect" gives Claude almost everything it needs to get the first draft right.
A useful pattern to keep in your head is role, context, task, format. Tell Claude what perspective to take (an editor, a skeptical reviewer, a specific kind of expert), give it the background it needs, state exactly what you want done, and specify what the output should look like, a table, a numbered list, a short paragraph, code with comments. You don't need all four in every prompt, but leaving out the format is the most common reason people get answers longer or shorter than they wanted.
It also helps to think of a chat as a conversation, not a single query. If the first answer is close but not quite right, don't start over, tell Claude specifically what to change ("shorter," "more casual," "cut the third paragraph," "assume the reader already knows the background"). Claude keeps the full thread in mind, so corrections compound rather than reset the conversation.
For anything you'll do more than once, a weekly report format, a specific editing style, a recurring analysis, it's worth setting that structure up once and reusing it, which is exactly what Projects are for.
Beyond a Single Chat
Projects: for work you come back to
A Project is a folder with memory: files and instructions that persist across every chat inside it.
Shared files
Upload reference material once, a style guide, a dataset, product documentation, and every chat in the project can draw on it without re-uploading.
Custom instructions
Set standing context Claude should apply to every conversation in the project, like a tone to keep, a role to play, or a format to default to.
Separate history
Each project keeps its own list of chats, so a project for client work doesn't mix with the chat where you were troubleshooting your resume.
Good for teams
If your plan includes shared projects, teammates working from the same brief get consistent output without everyone writing their own version of the same instructions.
What It Builds
Artifacts: when Claude gives you something to keep, not just read
When you ask for something like a document, a piece of code, a simple diagram, or a small interactive page, Claude often opens that output in a separate panel next to the chat rather than dropping it inline as chat text. That panel is called Artifacts, and the distinction matters practically: content in the chat log is conversational and hard to extract cleanly, while an Artifact is a standalone piece of content you can keep editing, view rendered (a webpage as a webpage, not as raw HTML text), and copy out in full.
This is especially useful for iterative work. If Claude writes a piece of code as an Artifact and you ask for a change, it updates that same artifact rather than making you scroll back through chat history to find the latest version. The same is true for a document draft, you can ask for edits in plain language and watch the artifact update, rather than regenerating the whole thing from scratch each time.
Not every response becomes an artifact. Short answers, explanations, and back-and-forth discussion stay in the regular chat. Claude generally reserves artifacts for content substantial or reusable enough to be worth extracting on its own.
Reading Your Material
What you can hand Claude directly
Claude isn't limited to what you type. You can bring your own material into the conversation.
Documents
PDFs, Word documents, and plain text files can be uploaded so Claude reads the actual content rather than relying on a description you type out.
Spreadsheets
CSV and spreadsheet files can be analyzed directly, Claude can summarize, spot patterns, or flag anomalies in the data rather than requiring you to paste rows manually.
Images
Photos, screenshots, and diagrams can be uploaded for Claude to describe, transcribe, or reason about, useful for reading a chart, debugging a UI screenshot, or transcribing handwritten notes.
Code and logs
Paste or upload source files, error logs, or config files directly for debugging help grounded in your actual code rather than a paraphrase of it.
Beyond the Browser
Where else Claude shows up
Claude.ai in the browser is the starting point, but it's not the only place Claude runs. Desktop and mobile apps mirror the same chats and projects so you're not locked to one device. For people who want Claude inside the tools they already use, there are browser extension and integration options that bring Claude into email, documents, and other everyday software rather than requiring a separate tab.
For anyone doing actual software development, Claude Code is a distinct product worth knowing about separately from Claude.ai. It runs in the terminal, works directly against your real codebase, reading files, running tests, making edits, using git, rather than the copy-paste-adapt loop of a chat window. If you're a developer and you've only used Claude through the chat interface, that's the biggest gap between what you've tried and what Claude can actually do for your work.
Which surface to use depends on the task, not a hierarchy, a browser chat is often the fastest way to draft an email, while a coding task with multiple files is a much better fit for Claude Code than pasting snippets back and forth.
Honest Limits
What Claude is not good at
Claude can state something confidently and still be wrong, particularly on specific facts like dates, statistics, names, or citations it hasn't been given directly. This isn't a rare glitch, it's a structural property of how these models generate text, and it means anything you plan to publish, act on financially, or rely on for a decision with real consequences should be independently checked, not taken on Claude's word alone.
Claude also doesn't have live awareness of the world beyond what you tell it or what a specific tool integration provides. Without a connected search or browsing feature, it's working from training data with a cutoff, so it can be behind on very recent events, current prices, or things that changed last week.
Long, highly technical, or deeply domain-specific tasks, a legal contract with jurisdiction-specific nuance, a medical question with real stakes, a niche academic citation, are places where Claude is a useful starting draft or second opinion, not a substitute for a qualified human. Treat it as a very capable collaborator that still needs a editor, not an authority that replaces one.
Finally, Claude works within a conversation's context window, it can lose track of very early details in an extremely long single chat. If a conversation has grown unwieldy, starting a fresh chat (inside the same Project, if relevant) and re-stating the essentials is often more reliable than continuing to scroll back.
Reading about Claude and being good at using it are different things
Everything on this page is faster to actually learn by doing it than by reading about it. The free Nightschool AI curriculum is hands-on from the first lesson, you'll be prompting, uploading files, and building real outputs in Claude, not just reading descriptions of what's possible.
Frequently Asked Questions
Do I need to know how to code to use Claude AI?
No. Claude.ai's chat interface is plain conversational text, you type what you want the way you'd explain it to a person. Coding knowledge only matters if you're using Claude Code, the separate developer tool for working inside real codebases.
Is Claude AI free to use?
Claude.ai offers a free tier you can sign up for and start using immediately. Paid plans exist for higher usage limits and access to more capable models, but you don't need one to try Claude and learn the basics.
What's the difference between a regular chat and a Project?
A regular chat is a single conversation that starts from scratch. A Project is a container for multiple chats that all share the same uploaded files and custom instructions, so you don't have to re-explain background information every time you start a new conversation on the same topic.
Can Claude read files I upload, like PDFs or spreadsheets?
Yes. You can attach documents, spreadsheets, and images directly into a chat, and Claude reads the actual content to answer questions, summarize, or analyze the data, it isn't limited to what you type.
What is the Artifacts panel for?
Artifacts is a side panel that opens when Claude produces standalone content like a document, code, or a simple webpage. It keeps that output separate from the conversational chat log so you can view it rendered, keep editing it, and export it cleanly.
Does Claude remember previous conversations?
Claude remembers everything within the current chat thread as you go. Whether it carries memory across separate chats depends on account and memory settings; Projects are the more reliable way to keep shared context available across multiple conversations.
Can Claude get things wrong?
Yes. Claude can state incorrect information confidently, especially specific facts like dates, statistics, or citations. Always verify anything you plan to publish, act on, or rely on for an important decision rather than trusting a single answer.
What's the difference between Claude.ai and Claude Code?
Claude.ai is the browser-based chat interface for general conversation, writing, and analysis. Claude Code is a separate terminal-based tool built for software development, it works directly against your real codebase, running commands and making file edits rather than requiring you to copy code back and forth from a chat.
Which Claude model should I use?
Claude.ai lets you pick a model from a dropdown in the chat interface. Faster, lighter models are a good fit for quick, simple tasks, while more capable models are worth choosing for harder reasoning, longer documents, or work where accuracy matters more than speed.
Can I use Claude outside the browser?
Yes. Claude is available through desktop and mobile apps that sync with your browser chats and projects, along with browser extension and integration options that bring Claude into other everyday tools like email and documents.
Stop reading about it. Start using it.
The fastest way to understand Claude is to open a chat and try one real task.