Claude: Beginner

Beginner

Claude is Anthropic’s chat app. It runs in your browser at claude.ai and as a native desktop app for macOS and Windows. You type, Claude answers, and you can attach files, images, and links for it to work with. This is the chat surface, separate from Claude Code, which runs in your terminal. This guide gets you from zero to a productive first chat.

New to AI tools in general? Skim How AI coding tools work first; it’ll make everything here click faster.

What you need

• A Claude account (the Free, Pro, or Max plan all work for this guide)
• A browser, or the desktop app installed
• Something real to work on: a document to summarise, an email to draft, a question to think through

Open Claude

1. Sign in or open the app:

   Web

   Go to claude.ai and sign in.

   Desktop

   Download the app for macOS or Windows, install it, and sign in with the same account.

2. Start a new chat and type a question in plain language:

   Explain what a context window is, in two sentences, for a non-technical colleague.

3. Read the answer, then reply in the same chat to refine it. Claude remembers everything earlier in the chat.

Tip

Claude is turn-based: you prompt, it answers, you steer the next turn. You’re never locked in. If an answer is off, say what’s wrong and ask again.

Write a prompt that works

The difference between a weak and a strong answer is almost always the prompt. Four habits:

1. Be specific. “Summarise this” is weak. “Summarise this contract in five bullets a non-lawyer can act on” gives Claude something to aim at.
2. Give context. Say who it’s for, what you already know, and what you’re trying to do.
3. Ask for a format. Bullets, a table, an email, 200 words: name it.
4. Iterate. Treat the first answer as a draft. “Shorter, and drop the jargon” gets you there faster than rewriting from scratch.

This is context engineering in miniature: the model has no opinion about direction; what you feed it is the steering.

Attach files for grounded answers

Instead of pasting long text, hand Claude the source directly. Use the attachment button in the message box to:

• Upload documents, spreadsheets, and images
• Paste a URL and ask Claude to read the page

Then ask questions against what you attached:

Here's our Q2 report (attached). What are the three numbers a board member would ask about?

Grounding answers in a real document is far more reliable than asking from memory. Everything you attach and type shares one context window, the model’s working memory for the chat.

Caution

Claude can make mistakes, especially with figures and quotes. Check anything load-bearing against the source before you act on it (see Verification).

Start fresh between topics

One chat should cover one task. When you switch to something unrelated, start a new chat. A focused conversation gives sharper answers than one long thread that drifts across five subjects.

You’re done when…

• ✅ You’ve signed in and held a back-and-forth chat
• ✅ You’ve attached a file and asked questions against it
• ✅ You know to start a new chat for a new topic

Next step

Ready for a real workflow with Projects that hold your context, custom instructions, and Memory?

Claude Intermediate Projects, custom instructions, and Memory: give Claude lasting context and keep each chat focused.