If you've just started learning to code, or you're a junior developer still getting comfortable with a terminal — a text-based window where you type commands instead of clicking icons — this guide is for you. It walks through installing Claude Code (Anthropic's coding agent, available in the terminal, code editors, and a desktop app), logging in, and making your first question, code change, and commit in a real project. Following it end to end takes about 15-20 minutes.
🟢 Model references match the current lineup · model notice · Fable subscription
Fable 5 and 5.1 subscription (updated September 7, 2026): Claude Fable 5.1, released September 1, 2026, is the current Fable model and Fable 5 is now legacy. Plan terms are the same for both — Max and Team Premium plans include Fable at up to 50% of the weekly usage limit; Pro and Team Standard use usage credits (
What do you need before you start?
You just need a computer where you can open a terminal, and a paid Claude subscription (or a Console account). Per the official requirements:
- Operating system — macOS 13.0+, Windows 10 version 1809+ (or Windows Server 2019+), Ubuntu 20.04+, Debian 10+, or Alpine Linux 3.19+
- Hardware — at least 4GB of RAM, an x64 or ARM64 processor
- Supported country — Claude Code is only available in countries Anthropic serves
- Internet connection — Claude Code needs to stay online to work
- A Claude subscription or Console account — Pro, Max, Team, or Enterprise, or a Claude Console account with prepaid API credits. The Free plan alone isn't enough to start Claude Code in the terminal
- (Optional) A project folder — not required. You can ask Claude Code to start a brand-new project for you
- (Windows only, optional) Git for Windows — lets Claude Code use its Bash tool. Without it, Claude Code falls back to PowerShell
Which terms should a beginner know first?
A handful of terms come up repeatedly below — defining them now avoids confusion later.
- Terminal — a text window where you type commands instead of clicking
- CLI (Command Line Interface) — a program you run by typing commands in a terminal. Claude Code runs as a CLI
- PATH — the list of folders your terminal searches to find a command like
claude. If the install folder isn't in it, you'll see a "command not found" error - CLAUDE.md — a memory file in your project folder that Claude Code reads at the start of every conversation, so it remembers project rules like build commands and code style without you repeating them
- Permission mode — the setting that controls whether Claude Code asks for your approval before editing files or running commands. There's an auto mode and a manual mode
- Plan mode — a mode where Claude reads code and drafts a plan without changing any files, useful for checking direction before you commit to an approach
- Context window — the amount of conversation and code Claude can hold in one session. Once it fills up, Claude is more likely to "forget" earlier instructions
How do you install Claude Code?
Paste one command into your terminal (or PowerShell) for your operating system, and the install finishes on its own.
macOS / Linux (in a terminal)
curl -fsSL https://claude.ai/install.sh | bash
Windows PowerShell (Win+X → Windows PowerShell)
irm https://claude.ai/install.ps1 | iex
If you're using Windows CMD instead
curl -fsSL https://claude.ai/install.cmd -o install.cmd && install.cmd && del install.cmd
When it's done, you'll see a completion message. To confirm, run:
claude --version
Success looks like this: a version number followed by (Claude Code).
Watch out: PowerShell and CMD look similar but use different commands. If your prompt starts with PS C:\, you're in PowerShell; if it's just C:\, you're in CMD. Running the wrong install command in the wrong shell fails immediately.
How do you log in and ask your first question?
After installing, move into your project folder and type claude — the login flow starts automatically.
cd /path/to/your/project
claude
On first run, a browser window opens asking you to sign in with your Claude account. Follow the on-screen steps. To switch accounts later, type /login inside a running session. Once you're logged in, you can ask about the project right away:
what does this project do?
what technologies does this project use?
where is the main entry point?
Claude Code reads whatever files it needs on its own — you don't need to paste in folder structure or file contents first.
How do you make your first code change and commit?
When you ask for a change, Claude Code finds the file and shows you what it plans to change. In manual mode it asks for approval before applying it.
add a hello world function to the main file
Select Yes to approve if it looks right. On Pro, Max, and Team plans, new sessions have started in auto mode by default since August 14, 2026. In auto mode a separate classifier model reviews each action in your place, flagging only the ones that look risky and letting the rest proceed without prompting you. If you set your own default mode in ~/.claude/settings.json, that setting is kept, and organization-managed defaults are not changed either. As of this writing, auto mode is still opt-in on Enterprise plans and via the Claude API, so those start in manual mode. Press Shift+Tab at any time to switch modes. Git operations work the same conversational way:
what files have I changed?
commit my changes with a descriptive message
Here are the commands you'll use most:
| Command | What it does | Example |
|---|---|---|
claude | Start an interactive session | claude |
claude "task" | Start a session with an initial request | claude "fix the build error" |
claude -p "query" | Ask once, then exit | claude -p "explain this function" |
/clear | Reset conversation history | /clear |
/help | Show available commands | /help |
The more specific your request, the fewer corrections you'll need afterward. The official guide's before/after examples show the difference clearly:
| Situation | Vague request | Specific request |
|---|---|---|
| Fixing a bug | "fix the login bug" | "users report that login fails after session timeout. Check the auth flow in src/auth/, especially token refresh. Write a failing test that reproduces the issue, then fix it" |
| Writing a test | "add tests for foo.py" | "write a test for foo.py covering the edge case where the user is logged out. Avoid mocks" |
| Build error | "the build is failing" | "the build fails with this error: [paste error]. Fix it and verify the build succeeds. Address the root cause, don't suppress the error" |
What order should you follow when adding a bigger feature?
The official guide recommends splitting anything beyond a small, clear fix into four phases: explore, plan, implement, commit. For a one-line typo fix or a small log statement, skip straight to asking — but for changes that touch multiple files or code you don't know well, this sequence keeps Claude from solving the wrong problem.
In the explore phase, press Shift+Tab until the status bar shows ⏸ plan mode on, then let Claude read files and answer questions without changing anything. In the plan phase, ask something like "I want to add this feature — what files need to change? Create a plan." Once you like the plan, approve it or press Shift+Tab to exit plan mode and let Claude implement it, then finish with "commit with a descriptive message and open a PR."
What are the most common install issues, and how do you fix them?
Most first-run errors trace back to PATH configuration or shell confusion. Here's what the official troubleshooting guide recommends:
| Symptom | Cause | Fix |
|---|---|---|
command not found: claude on macOS/Linux | The install folder isn't in your PATH | Add export PATH="$HOME/.local/bin:$PATH" to your shell config (~/.zshrc or ~/.bashrc) and open a new terminal |
'irm' is not recognized in PowerShell | You ran a PowerShell command in CMD | Open PowerShell (Win+X → Windows PowerShell) and retry, or use the CMD-specific install command instead |
claude is not recognized on Windows | The install folder isn't on your user PATH | Add %USERPROFILE%\.local\bin to your PATH from PowerShell, then open a new window |
syntax error near unexpected token '<' or HTML shows up in the terminal | The install URL returned an error response or a webpage instead of the script (the cause varies by situation) | Retry the command; if it persists, install with Homebrew instead (brew install --cask claude-code) |
Watch out: Claude Code's Bash tool needs Git for Windows. Without it, Claude Code falls back to PowerShell automatically — but if you see "Claude Code on Windows requires either Git for Windows or PowerShell," check whether PowerShell is on your PATH or whether Git is actually installed.
What should you try next?
Once the basic flow feels natural, run /init inside your project folder. It scans your project structure and drafts a starter CLAUDE.md file for you. Use that file for things Claude can't infer just by reading code — build commands, testing conventions, code style — so you don't have to re-explain them every session. If approval prompts feel repetitive, pre-approve commands you trust with /permissions. As you get more comfortable, try delegating research to a subagent (a helper agent that investigates in its own separate context) with a prompt like "use a subagent to investigate this" — it keeps your main conversation focused on implementation instead of exploration.
Frequently asked questions
Q. Can I use Claude Code on the Free plan?
No — logging into Claude Code in the terminal requires a Pro, Max, Team, or Enterprise subscription, or a Claude Console account. The Free plan alone won't get you started.
Q. Do I need to already know how to code?
No. The official guide is explicit that you don't need to know how to code — describe what you want in plain English (like "make me a simple webpage that says hello world") and Claude Code creates the files and asks for your approval.
Q. Should I use PowerShell or CMD on Windows?
Either works, but the commands differ. This guide's default install command is for PowerShell; if you're in CMD, use the separate CMD-specific command instead. Your prompt starts with PS C:\ when you're in PowerShell.
Q. Do I have to approve every single action?
It depends on your permission mode. On Pro, Max, and Team plans, new sessions have started in auto mode since August 14, 2026, so a classifier flags only risky-looking actions for approval. If you'd rather approve everything yourself, press Shift+Tab to switch to manual mode.
Related articles
- Claude Code Installation Guide: native vs npm, Windows Included
- Claude Skills for Automating Repetitive Work: Build Your Own Skill Without Coding
- Claude in Chrome for Office Work: Install, Set Up, and Automate Your First Task
- How to Use Claude Code: Real Usage After Install (Commands, VS Code)
- Claude Code Skills (Agent Skills) Guide: Extend Claude with SKILL.md