Complete Claude Code CLI Reference

Every command, flag, keyboard shortcut, and interactive feature in Claude Code.

CLI Commands

Core commands to interact with Claude Code from the terminal.

Interactive Session

claude — Start an interactive conversation

One-Shot Query

claude "your question" — Quick inline query

Print Mode (Non-Interactive)

claude -p "query" — Output response and exit. Perfect for scripts and piping.

Interactive Mode

Keyboard shortcuts and interactive features available during a Claude session.

Plan Mode (Shift+Tab)

Toggle Plan Mode to make Claude think and plan before writing code. Claude explores, analyzes, and proposes a plan for your approval.

Voice Dictation & Bash Mode

Hold Space for voice dictation. Type ! for inline shell commands. Use @ for file references.

Keyboard Shortcuts

Ctrl+C: Cancel. Ctrl+G: Fuzzy search. Ctrl+B: Background agent. Ctrl+T: Task list. Shift+Tab: Plan Mode.

Advanced Features

Power features for advanced workflows and automation.

Extended Thinking

Use 'think', 'think hard', 'think harder', or 'ultrathink' to activate deeper reasoning. More thinking budget = more thorough analysis for complex problems.

Project Memory (CLAUDE.md)

Create CLAUDE.md files to store project rules, conventions, and context. Claude reads these automatically. Keep them short, specific, and regularly pruned.

Checkpoints & Rewind

Claude automatically creates checkpoints. Use /rewind to go back to any previous state if something goes wrong.

NEW in 2.1

Latest in Claude Code

⚡Agent Teams

Run parallel agent teams with shared task lists and mailboxes

🔀Background Subagents

Run agents in background with Ctrl+B, check results later

🎯MCP Tool Search

Auto-deferred MCP tools save context window space

⌨️Custom Keybindings

Customize all keyboard shortcuts via /keybindings

🌐Desktop & Web

Desktop App for macOS/Windows and Web at claude.ai/code

📋Worktree Isolation

Run agents in isolated git worktrees with --worktree flag

CLI Flags & Options

Comprehensive list of CLI flags and options for customizing Claude Code behavior.

Model & Execution Flags

--model (-m): Select model. --effort: Set thinking effort. --max-turns: Limit turns. --max-budget-usd: Set spending cap. --agent: Use a subagent.

System Prompt & Output Flags

--system-prompt: Override prompt. --output-format (text/json/stream-json): Control output. --json-schema: Validate JSON.

Tool & Session Flags

--allowedTools/--disallowedTools: Control tools. --worktree: Git isolation. --chrome: Chrome integration. --plugin-dir: Load plugins.

Practical Command Examples

Exploration

Codebase Overview

claude> analyze the overall structure of this project

Find Features

claude> document the API endpoints

Development

Build Features

claude> add user profile editing functionality

Test & Debug

claude> run all tests and analyze failure causes

CLI Tips

Be Specific

Instead of 'fix the error', say 'fix the null reference error in the handleSubmit function in user.tsx'.

Use Plan Mode First

Press Shift+Tab to enable Plan Mode. Let Claude explore and plan before making changes.

Manage Context

Use /clear to start fresh, /compact to compress, and /btw for side questions. Keep CLAUDE.md short and specific.