Claude Code CLI Reference

Complete reference for the Claude Code command-line interface.

Installation

System Requirements

• macOS 13.0+, Windows 10 1809+, Ubuntu 20.04+, Debian 10+, Alpine 3.19+
• 4GB+ RAM
• Internet connection required
• Shell: Bash, Zsh, PowerShell, CMD
• Windows: Git for Windows required

Methods

# Native (Recommended)
curl -fsSL https://claude.ai/install.sh | bash

# Homebrew (macOS)
brew install --cask claude-code

# WinGet (Windows)
winget install Anthropic.ClaudeCode

# npm (legacy)
npm install -g @anthropic-ai/claude-code

# Direct invocation
npx @anthropic-ai/claude-code

# Update to latest
npm update -g @anthropic-ai/claude-code

Initial Login

claude     # First run opens browser for login

Supports: Claude Pro, Teams, Enterprise, Console, Bedrock, Vertex AI, Foundry. Credentials stored at ~/.claude.json.

Basic Usage

# Interactive mode (default)
claude

# Start with initial prompt
claude "explain this codebase"

# Non-interactive / print mode
claude -p "what does this function do?"

# Pipe input
cat file.ts | claude -p "review this code"

# Resume last conversation
claude --continue

# Resume specific conversation
claude --resume <session-id>

CLI Flags

Session Management Flags

Flag	Short	Description
--help	-h	Show help message
--version	-v	Show version number
--print	-p	Non-interactive mode: print response and exit
--verbose		Enable verbose/debug logging
--debug [category]		Debug logging for specific category
--model	-m	Override model (e.g., sonnet, opus, haiku)
--continue	-c	Resume most recent conversation
--resume	-r	Resume specific conversation by ID
--fork-session		Create new session from current context
--teleport		Resume a web session locally
--remote "task"		Create cloud session
--worktree	-w	Run in isolated git worktree

Output & Format Flags

Flag	Description
--output-format	Output format: text (default), json, stream-json
--max-turns	Maximum conversation turns in non-interactive mode
--max-budget-usd	Spending limit per session
--no-input	Skip initial user input (for piped content)
--input-format	Input format: text (default), stream-json
--json-schema	Structured output with JSON schema (print mode)

Permission & Security Flags

Flag	Description
--dangerously-skip-permissions	Skip all permission prompts (use with caution)
--allowedTools	Comma-separated list of allowed tools
--disallowedTools	Comma-separated list of disallowed tools
--permission-mode	Permission mode: default, plan, acceptEdits, dontAsk, bypassPermissions
--permission-prompt-tool	MCP tool for handling permission prompts

Configuration Flags

Flag	Description
--system-prompt	Override entire system prompt (use with -p)
--append-system-prompt	Append text to system prompt
--system-prompt-file	Load system prompt from file
--append-system-prompt-file	Append system prompt from file
--mcp-config	Path to MCP config JSON file
--settings	Load settings from file or JSON string
--agents	Define custom agents as JSON
--plugin-dir	Load plugin from directory (repeatable)
--add-dir	Add additional working directories to context
--init	Run initialization hooks
--maintenance	Run maintenance hooks

PR & Review Flags

Flag	Description
--from-pr <url>	Load PR context from GitHub URL
--fallback-model	Fallback model if primary unavailable

Agent Team Flags

Flag	Description
--teammate-mode	Run as part of an agent team
--chrome	Enable Chrome browser for agent
--no-chrome	Disable Chrome browser
--strict-mcp-config	Fail on invalid MCP config
--setting-sources	Show which file each setting comes from
--include-partial-messages	Include partial messages in stream output
--betas	Enable beta features
--no-session-persistence	Disable session persistence
--allow-dangerously-skip-permissions	Allow skipping permissions programmatically

Auth CLI

claude auth login              # Login to Anthropic
claude auth logout             # Logout
claude auth status             # Show auth status

claude agents                  # Manage agent teams
claude update                  # Update Claude Code

claude mcp add-json <name> <json>   # Add MCP from JSON blob
claude mcp add-from-claude-desktop  # Import from Claude Desktop

Session ID Flags

Flag	Description
--session-id	Set specific session ID
--conversation-id	Set conversation ID within session

Environment Variables

Authentication

Variable	Description
ANTHROPIC_API_KEY	API key for Anthropic direct access
ANTHROPIC_AUTH_TOKEN	OAuth/bearer token for API access
CLAUDE_CODE_API_KEY	Alternative API key variable

Provider Selection

Variable	Description
CLAUDE_CODE_USE_BEDROCK	1 to use AWS Bedrock
CLAUDE_CODE_USE_VERTEX	1 to use Google Vertex AI
ANTHROPIC_BASE_URL	Custom API base URL

AWS Bedrock Configuration

Variable	Description
ANTHROPIC_BEDROCK_BASE_URL	Bedrock endpoint URL
AWS_REGION	AWS region for Bedrock
AWS_ACCESS_KEY_ID	AWS access key
AWS_SECRET_ACCESS_KEY	AWS secret key
AWS_SESSION_TOKEN	AWS session token (for temporary creds)
AWS_PROFILE	AWS profile name
ANTHROPIC_MODEL	Override model ID for Bedrock

Google Vertex Configuration

Variable	Description
ANTHROPIC_VERTEX_PROJECT_ID	GCP project ID
CLOUD_ML_REGION	GCP region
ANTHROPIC_VERTEX_BASE_URL	Custom Vertex endpoint

Model Overrides

Variable	Description
ANTHROPIC_MODEL	Override default model ID
ANTHROPIC_DEFAULT_SONNET_MODEL	Override Sonnet model ID
ANTHROPIC_DEFAULT_OPUS_MODEL	Override Opus model ID
CLAUDE_CODE_EFFORT_LEVEL	Reasoning effort: low, medium, high

Behavior Configuration

Variable	Description
DISABLE_AUTOMEMORY	1 to disable auto-memory
DISABLE_AUTOUPDATER	1 to disable auto-updates
DISABLE_BUG_COMMAND	1 to disable /bug command
CLAUDE_CODE_MAX_OUTPUT_TOKENS	Override max output tokens
CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC	1 to disable telemetry
CLAUDE_CODE_ENABLE_TELEMETRY	Enable OpenTelemetry
CLAUDE_CODE_SKIP_DOCTOR	1 to skip doctor checks
CLAUDE_CODE_DISABLE_FAST_MODE	Disable fast Opus mode
CLAUDE_CODE_SIMPLE	Use minimal system prompt
CLAUDE_CODE_DISABLE_1M_CONTEXT	Disable large context window
CLAUDE_CODE_SHELL	Override shell detection
BASH_DEFAULT_TIMEOUT_MS	Bash command timeout in ms
MAX_MCP_OUTPUT_TOKENS	MCP tool output token limit
SLASH_COMMAND_TOOL_CHAR_BUDGET	Skill character budget

Directory Overrides

Variable	Description
CLAUDE_CONFIG_DIR	Custom config directory location
CLAUDE_CODE_TMPDIR	Override temporary directory

Cloud Provider: Microsoft Foundry

Variable	Description
CLAUDE_CODE_USE_FOUNDRY	1 to use Microsoft Foundry

Agent Teams & Advanced

Variable	Description
CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS	Enable agent teams feature
CLAUDE_CODE_TEAM_NAME	Team name for agent teams
CLAUDE_CODE_PLAN_MODE_REQUIRED	Require plan mode for teammates
CLAUDE_CODE_DISABLE_CRON	Disable background cron tasks
CLAUDE_CODE_DISABLE_BACKGROUND_TASKS	Disable all background tasks
CLAUDE_CODE_SHELL_PREFIX	Prefix for shell commands

mTLS & Client Certificates

Variable	Description
CLAUDE_CODE_CLIENT_CERT	Client certificate path for mTLS
CLAUDE_CODE_CLIENT_KEY	Client private key path
CLAUDE_CODE_CLIENT_KEY_PASSPHRASE	Passphrase for client key

Proxy & Network

Variable	Description
HTTP_PROXY / HTTPS_PROXY	HTTP/S proxy URL
NO_PROXY	Proxy bypass list
NODE_TLS_REJECT_UNAUTHORIZED	0 to skip TLS verification
CLAUDE_CODE_PROXY_RESOLVES_HOSTS	Let proxy handle DNS resolution

Model Aliases

Alias	Resolves To
opus	claude-opus-4-6
sonnet	claude-sonnet-4-6
haiku	claude-haiku-4-5-20251001
opusplan	Opus with plan mode
sonnet[1m]	Sonnet with 1M context
claude-opus-4-6[1m]	Opus with 1M context

Exit Codes

Code	Meaning
0	Success
1	General error
2	Invalid arguments

Examples

# Use specific model
claude -m claude-haiku-4-5-20251001 -p "quick question"

# JSON output for scripting
claude -p "list files" --output-format json

# Multi-turn with max turns
claude -p "refactor this" --max-turns 5

# With Bedrock
CLAUDE_CODE_USE_BEDROCK=1 AWS_REGION=us-east-1 claude

# With Vertex
CLAUDE_CODE_USE_VERTEX=1 ANTHROPIC_VERTEX_PROJECT_ID=my-project claude

# Add extra directories
claude --add-dir /path/to/other/repo

# With custom MCP config
claude --mcp-config ./custom-mcp.json