You have installed Claude Code. The terminal is open or the VS Code extension is installed and authenticated. You are looking at a Claude Code session and you are not sure what to do first.
This guide takes you through a structured 30 minutes that turns a bare install into a session you actually accomplished something with. No theory. Just a runnable script you follow step by step.
Before you start: pick a folder
Claude Code works inside a folder of your choosing — your workspace. For this first session, use a folder that has some real content but is not critical work: a folder with notes, drafts, downloaded PDFs, or a project you have been meaning to organise. Avoid system folders or folders containing sensitive material you have not yet decided whether to share with Claude.
If you do not have a suitable folder, create a new one called ~/claude-test and copy two or three files into it (any files — Word docs, text files, PDFs, an old report). That is enough to make the session real.
Minute 0–3: Open Claude Code in that folder
Terminal route: open Terminal (or your shell), then:
cd ~/your-folder
claudeVS Code route: open VS Code, then File → Open Folder and select your folder. Click the Claude Code icon in the left activity bar to open the panel. Sign in if prompted.
Either way, you should now see a Claude Code prompt waiting for input. The folder you opened it in is now Claude's workspace for this session — every file inside is visible to Claude.
Minute 3–8: Ask Claude what is in there
Type:
What files are in this folder, and what is each one about? Just give me a one-line description of each.
Claude will use its file tools to list the files, read enough of each to understand it, and reply with a short description per file. This proves three things at once: Claude can see your files, Claude can read them without you having to upload anything, and Claude will tell you what it found.
If Claude reads a sensitive file you did not want it to see, that is the moment to learn that the workspace boundary is set by which folder you opened. Anything inside is in scope; anything outside is not.
Minute 8–13: Ask a real question about a real file
Pick one specific file from the list and ask Claude something useful about it. Examples:
- "Read [filename]. What are the three most important points?"
- "Look at [filename] and tell me what is missing or unclear."
- "Summarise [filename] for someone who has 30 seconds."
- "Find every place in [filename] that mentions [topic] and quote it."
The point is to use Claude on something you actually care about, not on a fake demo. Read the response. If it is useful, ask a follow-up. If not, refine the prompt — see How to Talk to Claude.
Minute 13–18: Make a small edit
Now do something Claude Code can do that Desktop cannot: have Claude edit a file directly. Pick a low-stakes file (or copy one to a temporary name first if you want a safety net).
Try one of:
- "In [filename], change the date in the header to today's date and save."
- "In [filename], rewrite the first paragraph to be tighter and clearer. Keep everything else unchanged."
- "In [filename], find any spelling errors and fix them. Show me the changes before saving."
Claude will show you a diff — a side-by-side view of what would change — and ask you to confirm before writing to disk. Read the diff. Approve if it looks right; reject and refine if not. This permission flow is the core safety model of Claude Code: you see every change before it happens.
Once you approve and the edit is written, open the file in your normal editor to confirm the change is real. It is.
Minute 18–25: Create a CLAUDE.md so the next session starts smarter
Right now, every time you open Claude Code in this folder, it knows nothing about you or this folder. Fix that with a CLAUDE.md file.
Ask Claude:
Create a CLAUDE.md file in this folder. Include: a short description of what this folder is for, the conventions I want followed (UK English, professional tone, no emojis), and a list of files in here that you found useful for context. Keep it under 60 lines.
Claude will draft a CLAUDE.md based on what it learned in the previous steps and propose it as a new file. Approve. Open it in any text editor afterwards and adjust anything that does not match how you actually work.
From the next Claude Code session in this folder onwards, CLAUDE.md is read automatically before the first message — Claude knows the context without you having to repeat it.
Minute 25–30: End the session cleanly
Two things to do before you close the terminal or VS Code:
1. Tell Claude what to remember for next time. Type:
Remember that I prefer [your preference here — e.g. concise responses, no preamble, UK English]. Add anything else useful you have learned about how I work in this session.
Claude will write this to its auto memory file. Next session, it loads automatically.
2. Close the session. In the terminal, press Ctrl+C twice or type /exit. In VS Code, just close the panel. Your changes are already on disk; nothing else needs saving.
What you accomplished
In 30 minutes you have: opened a workspace, given Claude visibility of your files, asked questions about real content, made and approved a real file edit, created a CLAUDE.md so the next session inherits the context, and stored a memory entry so Claude remembers your preferences. That is the full Claude Code loop. Everything else in this hub is layered on top of those primitives.
What comes next
- Your First Claude Code Workspace — go deeper on CLAUDE.md, settings.json, and the rules folder
- Memory in Claude Code — the two memory systems and how they fit together
- Installing Claude Code — if anything in your install needs revisiting
- Common Mistakes Beginners Make — read this before you go further; cheaper than learning the hard way