How to Use Claude Code: A Beginner's First Week

Learning how to use Claude Code takes one afternoon for the basics and about a week for the habits that make it reliable. You install it, open a project folder, and describe what you want in plain language. Claude reads your files, proposes changes and runs commands with your permission.

The mistake most beginners make is jumping straight to "build me an app." What works better is a first week that moves from questions, to small changes, to written rules, to bigger tasks. That's the order I use when I teach it, and it's the plan below.

What do you need before you start?#

Three things, according to the official Claude Code quickstart:

• A terminal. Terminal on macOS, PowerShell on Windows, or any Linux shell.
• A project to work on. A code repository, but a folder of documents works too.
• An account. A paid Claude plan (Pro, Max, Team or Enterprise), a Claude Console account with API credits, or access through a cloud provider.

Claude Code isn't on the free plan. If you're deciding which plan to buy, read my breakdown of Claude Code pricing first.

Day 1: Install Claude Code and explore a project#

Install it with the native installer, which updates itself in the background. On macOS, Linux or WSL:

curl -fsSL https://claude.ai/install.sh | bash

On Windows, in PowerShell:

irm https://claude.ai/install.ps1 | iex

Prefer a package manager? brew install --cask claude-code on macOS and winget install Anthropic.ClaudeCode on Windows both work, but they don't auto-update. On Windows, install Git for Windows as well so Claude Code can use its Bash tool.

Check it worked with claude --version, then open a project and start a session:

cd path/to/your/project
claude

The first run opens your browser to log in. After that, spend day one only asking questions:

• what does this project do?
• where is the main entry point?
• explain the folder structure
• what would you change first, and why?

You don't need to paste files in. Claude Code reads what it needs as it goes, which is the first thing that feels different from a chat window.

Day 2: Make your first change safely#

Now ask for something small and specific. "Fix the bug" is too vague; "users see a blank screen after entering a wrong password on the login page, find out why and fix it" gives Claude something it can verify.

Before any change, create a branch so everything is easy to undo:

git checkout -b claude/first-change

Claude shows each proposed edit. On Pro, Max and Team plans, interactive sessions start in auto mode, where a classifier reviews actions so Claude can edit most files without stopping to ask you. Press Shift+Tab at any time to cycle permission modes if you want to approve every step while you're learning.

Git itself becomes conversational: ask "what files have I changed?" or "commit these changes with a descriptive message."

Day 3: Write a CLAUDE.md file#

By day three you'll notice you keep repeating yourself: which package manager to use, how to run tests, where things live. That belongs in a CLAUDE.md file, which Claude reads at the start of every session.

Run /init and Claude will generate a starting file from your codebase. Then trim and sharpen it. The memory documentation recommends keeping each file under 200 lines and writing instructions you could verify, such as "run npm test before committing" instead of "test your changes."

I wrote a full guide with a copyable template: how to write a CLAUDE.md Claude actually follows.

Day 4: Plan before bigger tasks#

For anything that touches several files, switch to plan mode first. Press Shift+Tab until you reach it, describe the goal, and Claude explores the codebase and proposes an approach before editing anything.

Read the plan like a code review. Push back on anything that looks too broad, ask for smaller steps, and only then let it implement. This single habit saves more wasted sessions than any prompt trick.

When Claude heads in the wrong direction mid-task, press Escape to stop it. /rewind, or pressing Escape twice, restores the conversation and code to an earlier checkpoint.

Day 5: Learn the commands you'll use daily#

You only need a handful:

Type / inside a session to see commands and skills, and use Tab to autocomplete.

Day 6: Add a skill or an MCP server#

Once the basics feel natural, extend Claude Code in one of two directions.

Skills package a repeatable workflow, such as your code review checklist or a release routine, into a folder Claude loads only when the task matches. My guide to Claude skills walks through writing your first one.

MCP servers connect Claude Code to outside tools like GitHub, a database or your Obsidian vault. Start with one server you'd genuinely use; every connected server adds context to your sessions.

Day 7: Review your week and your costs#

End the week by running /usage to see what consumed your plan, and /insights for a report on how you work and where sessions went wrong. Anthropic's cost guide lists the habits that matter most: clear between tasks, default to Sonnet, and keep instructions short.

Then write down three tasks you repeat every week. Those are next week's candidates for a skill, a hook or a scheduled job.

Where to go from here#

After this first week you'll know enough to be dangerous and not enough to be fast. The gap is mostly practice on real work: sharper task descriptions, better project rules and knowing when to delegate to a subagent.

If you want to close that gap with feedback on your own repository, my four-week Claude Code course picks up exactly where this plan ends. If you're still choosing tools, compare Cursor vs Claude Code before you commit.