AI

Claude CLI: The Complete Guide to the Command Line AI Assistant

Amine LahmaryAmine Lahmary·Aug 3, 2026·5 min read
Claude CLI: The Complete Guide to the Command Line AI Assistant

What is the Claude CLI?

The Claude CLI is a command-line tool made by Anthropic. You start it in any project folder, and Claude can:

  • Read and understand your whole codebase.
  • Edit files and write new code.
  • Run tests and commands for you.
  • Explain bugs, review changes, and suggest fixes.
  • Work with git, tools, and APIs. It works in three ways:
  1. Interactive mode — a chat-like session in your terminal.
  2. One-time queries — ask a question and get an answer, then exit.
  3. Piped input — feed it a file or log to analyze.

    How to install the Claude CLI

Installation is quick. Choose the method for your system.

Windows

Open PowerShell and run:

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

If you prefer the Windows package manager:

winget install Anthropic.ClaudeCode

macOS and Linux

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

On macOS you can also use Homebrew:

brew install --cask claude-code

With npm

If you already use Node.js:

npm install -g @anthropic-ai/claude-code

Check the installation

claude --version

You should see a version number followed by (Claude Code).

Log in and start your first session

The Claude CLI needs an account. Open your terminal, go to your project folder, and run:

claude

On first use, Claude will ask you to log in. After that, you can start working immediately.

Tip: Install Git for Windows on native Windows so Claude can use the Bash tool. Otherwise, Claude automatically uses PowerShell.

Essential Claude CLI commands

Here are the commands you will use every day.

Start and resume sessions

CommandWhat it doesclaudeStart an interactive sessionclaude "fix the build error"Start a session with a taskclaude -p "explain this function"Run a one-time query, then exitclaude -cContinue the most recent conversationclaude -rResume a previous conversationcat logs.txt | claude -p "explain"Analyze a file with Claude

The -p (print) flag is perfect for quick questions. Use -c to continue yesterday's work, and -r to pick a specific old session.

Commands inside a session

Type these with a leading slash inside Claude Code:

CommandWhat it does/helpShow all available commands/clearClear the conversation history/initCreate a CLAUDE.md memory file/statusShow your current session info/modelSwitch the AI model/compactSummarize a long conversation to save context/exitLeave Claude Code

Useful Claude CLI flags

Flags change how Claude starts. These are the ones you will use most.

--model

Choose which model to use for the session:

claude --model claude-sonnet-5

You can also use short names like sonnet, opus, or haiku.

--permission-mode

Control how much Claude can do without asking:

claude --permission-mode plan
claude --permission-mode acceptEdits

plan is great for a first pass: Claude explains what it will do without changing files.

--dangerously-skip-permissions

Runs Claude without asking for permission on every action. Use it only in safe environments, like a sandbox or CI:

claude --dangerously-skip-permissions "refactor the login component"

--add-dir

Let Claude read and edit code from other folders:

claude --add-dir ../shared-components

--bg(background)

Start a session in the background and keep working:

claude --bg "investigate the flaky test"

Claude gives you a session ID so you can check the result later.

Set up your project memory

Claude can remember your project's rules. Run /init inside a session and Claude creates a CLAUDE.md file. Use it to store:

  • The commands to run tests and builds.
  • Your coding style and conventions.
  • The project structure.
  • Common pitfalls to avoid. A good CLAUDE.md makes every future session faster and more accurate.

Daily workflow tips

Here is a simple workflow that works well for most developers:

  1. Plan first. Start with claude --permission-mode plan and ask Claude to explain its approach before it touches code.
  2. Give one clear task. Instead of "fix the app", say "fix the login bug where the form submits before validation".
  3. Continue, don't restart. Use claude -c to keep context between tasks in the same project.
  4. Use print mode for quick questions. claude -p "what does this function do?" is faster than opening a full session.
  5. Review before you trust. Always read the diff. Claude is fast, but you are responsible.
  6. Add memory. Keep CLAUDE.md updated so Claude improves over time.

    Extend Claude with skills and tools

Once you are comfortable, Claude Code gets even more powerful:

  • Skills — folders with a SKILL.md file that teach Claude your team's patterns. Run them with /skill-name.
  • Subagents — specialized assistants for tasks like code review.
  • MCP servers — connect Claude to databases, browsers, or your own APIs.
  • Hooks — run scripts automatically when Claude performs certain actions. Start with the essentials first. The power comes later, once the basics feel natural.

Troubleshooting quick tips

  • claudeis not recognized → close and reopen your terminal, or reinstall with the installer above.
  • Login problems → run claude auth login again.
  • Slow or long sessions → use /compact to summarize the conversation and free up space.
  • Wrong project rules → edit your CLAUDE.md file and restart the session.

    Start using the Claude CLI today

The Claude CLI turns your terminal into a coding partner. Install it in five minutes, learn the essential commands, and let Claude handle the repetitive work while you focus on the hard problems.

Try this right now: open your project, run claude, and ask it to explain your codebase. You will see the power of an AI assistant that lives where you work.

This post was written in August 2026 based on the latest Claude Code documentation. Features and commands evolve quickly, so check the official docs for updates.