Claude Code · Terminal Setup Guide
Claude in your terminal.
A practical guide to installing Claude Code, opening the right project folder, understanding what it can access and starting your first real AI-assisted development session.
PS C:\Projects\my-site> claude
✦ Welcome to Claude Code
Working directory:
C:\Projects\my-site
Claude: What would you like to work on?
> Explain this project and show me how to run it locally.
Before installing anything
The folder where you start Claude becomes its working environment.
Claude Code is not simply a chatbot inside a black terminal window. It is an AI coding agent that can read a project, search through its files, edit code and run development commands.
When you open a terminal, move into a project folder and run claude, that folder becomes the center of the session. Claude can understand the project as a connected system instead of seeing only the single file you paste into a chat.
This is why opening the correct directory matters. If your website is inside a folder called my-site, start Claude inside that folder—not from your entire drive or home directory.
The mental model
Three parts make the whole system work.
The doorway
PowerShell, Terminal or another command-line application is where you launch Claude and where Claude runs project commands.
The context
The active folder contains the code, assets, configuration and instructions Claude uses to understand what you are building.
The boundary
Claude can propose edits and commands, but permission settings determine which actions happen automatically and which require your approval.
The important shift
Stop sending files one by one.
Give the agent a real project environment and it can follow relationships between files, run the application and verify its own work.
Installation
From an empty terminal to your first Claude session.
Prepare your machine
You need a supported Claude account, an internet connection and a terminal. Claude Code supports current versions of Windows, macOS and several Linux distributions.
On Windows, installing Git for Windows is recommended. It gives Claude access to Git and a Bash environment. Claude can also use PowerShell when Git Bash is unavailable.
Node.js is useful for JavaScript and React projects, but the current native Claude Code installer does not require a global npm installation.
Install Claude Code
Choose the command for your operating system. These commands use Anthropic’s official native installer.
If Windows shows an irm is not recognized message, check that you opened PowerShell rather than Command Prompt.
Verify the installation
Close and reopen the terminal so the updated command path is available, then check the installation.
claude doctor performs a read-only health check and reports installation, update or configuration problems.
Open your project folder
Use cd to move into the folder you want Claude to understand. Quotation marks are useful when a folder name contains spaces.
Sign in and begin
The first launch opens an authentication flow. Sign in using your eligible Claude subscription or Anthropic Console account, then return to the terminal.
Access and permissions
Claude does not automatically receive unlimited access to your computer.
Project access
Claude starts with the folder where you launched it. It can inspect the files and relationships inside that workspace.
Permission prompts
Editing files and running some terminal commands may require your approval. Read the requested action before accepting it.
Explicit access
When a project depends on another directory, add it intentionally instead of starting Claude from a huge parent folder.
Start inside the smallest folder that contains the project. This gives Claude focused context, reduces unnecessary searching and avoids exposing unrelated personal files.
Your first real workflow
Treat Claude like an engineer working inside the project.
“Inspect this codebase and explain the architecture in simple language. Do not change anything yet.”
“Find the correct development command, start the application and tell me where I can preview it.”
“Reproduce this error, identify the root cause and explain your proposed fix before editing the code.”
“Create a short implementation plan for this feature. Preserve the existing visual style and verify the result after building it.”
Project memory
Give Claude instructions that survive every new session.
A CLAUDE.md file lives inside your project and explains how Claude should work there.
Add the project purpose, important commands, design rules and things Claude must never change. Claude reads these instructions when a new session begins.
You can ask Claude to create a starting version by running /init inside the session.
Useful commands
The small command list worth remembering.
claude
Start an interactive Claude Code session in the current folder.
claude -c
Continue the most recent conversation for the current project.
claude --resume
Choose and reopen an earlier Claude Code session.
claude doctor
Check the installation and configuration without starting a session.
/init
Create initial project instructions for Claude inside CLAUDE.md.
/help
Display the commands available in your current Claude Code version.
/permissions
Review or update the actions Claude is allowed to perform.
/context
See what is using space in the current conversation context.
/compact
Summarize a long conversation and free space for continued work.
The terminal is the interface. The project is the context. Git is the safety net.
The final thought
Start small. Stay in control. Let the agent earn more trust.
Your first Claude Code session does not need an advanced automation system, dozens of tools or a perfect configuration.
Start inside one real project. Ask Claude to inspect it. Review the plan, approve changes carefully and check the result.
Once that basic loop becomes familiar, the terminal stops feeling like a technical barrier. It becomes a direct interface between your idea and the system capable of building it.