Reference

Claude Code cheatsheet

Quick reference card for Claude Code: all commands, shortcuts, and tips in one page.

Publié le 11 mars 2026

Slash commands

Type these commands directly in the Claude Code prompt (interactive session).

Command	Description
`/help`	Show help and list available commands
`/clear`	Clear the current conversation history
`/compact`	Compress context (automatic summary of previous exchanges)
`/cost`	Show the token cost of the current session
`/doctor`	System diagnostics: network, MCP, permissions, version
`/init`	Generate a `CLAUDE.md` file in the current directory
`/review`	Request a code review on modified files
`/memory`	View or edit persistent memory (active CLAUDE.md file)
`/model`	View or change the current model (`haiku`, `sonnet`, `opus`)
`/permissions`	View and manage permissions granted to tools
`/bug`	Open a bug report to Anthropic
`/config`	Show the active configuration (merged settings.json)
`/login`	Authenticate via browser (Claude.ai Max)
`/logout`	Log out from your Claude.ai account
`/status`	Show account status, quotas, and connected MCPs
`/mcp`	Manage MCPs: list, enable, disable
`/install-github`	Install the official GitHub MCP
`/terminal-setup`	Configure terminal integration (shell hooks)
`/vim`	Enable vi/vim input mode in the terminal
`/resume`	Resume the last interrupted session

Slash command autocompletion

In an interactive session, type / then press Tab to see all available commands with a short description.

Keyboard shortcuts

Shortcut	Action
`Escape`	Cancel the current generation (without quitting)
`Ctrl+C`	Interrupt generation or exit the current prompt
`Ctrl+D`	Quit Claude Code cleanly
`Ctrl+L`	Clear the terminal display (not the context)
`Shift+Tab`	Insert a line break in the prompt (multiline)
`Alt+T` / `Option+T`	Toggle Extended Thinking on/off
`Ctrl+O`	Show reasoning details (verbose thinking)
`Up Arrow`	Recall the previous command from history
`Down Arrow`	Move forward through command history
`Tab`	Autocompletion (files, slash commands)

OS-specific shortcuts

Alt+T works on Windows/Linux. Option+T is the macOS equivalent. Both toggle Extended Thinking on or off.

Configuration files

Files are read in this order, from highest to lowest priority.

File	Scope	Description
`.claude/settings.json`	Project	Project-local configuration (committed or not)
`.claude/settings.local.json`	Project (local)	Local overrides, not committed (gitignored)
`~/.claude/settings.json`	User	Global configuration for all projects
`CLAUDE.md`	Project	Natural language instructions read at startup
`.claude/CLAUDE.md`	Project (hidden)	Alternative variant of the CLAUDE.md file
`~/.claude/CLAUDE.md`	Global	Global instructions applied to all projects
`.claude/agents/`	Project	`.md` files defining custom agents
`~/.claude/agents/`	Global	Global agents available across all projects
`.claude/skills/`	Project	Custom Skills (slash commands) for the project
`~/.claude/skills/`	Global	Global Skills available everywhere

3 execution modes

Interactive mode (default)

# Launch an interactive REPL session
claude

# Launch with a specific model
claude --model claude-sonnet-4-5

Used for day-to-day development. Claude maintains context between exchanges.

Print mode (`--print` / `-p`)

# Execute a task, display the result, then quit
claude --print "Explain what this file does"

# Short alias
claude -p "Summarize the contents of README.md"

# With JSON format
claude --print --output-format json "List the functions in src/utils.ts"

# Pipe mode (stdin -> stdout)
echo "Fix the typos in this text" | claude --print
cat file.ts | claude --print "Add JSDoc comments"

Headless / SDK mode

# For automation and CI/CD
claude --print --max-turns 1 --output-format stream-json "prompt"

# Environment variables for non-interactive mode
ANTHROPIC_API_KEY=sk-ant-... claude --print "prompt"

MCP commands

Command	Description
`claude mcp add <name> -- <cmd> <args>`	Add an MCP via stdio (global scope by default)
`claude mcp add --scope project <name> -- <cmd>`	Add an MCP at project scope
`claude mcp add --transport sse <name> <url>`	Add an MCP via SSE transport
`claude mcp remove <name>`	Remove a configured MCP
`claude mcp list`	List all configured MCPs with their status
`claude mcp logs <name>`	Show logs for a specific MCP
`claude mcp get <name>`	Show detailed configuration for an MCP

# Practical examples
claude mcp add filesystem -- npx -y @modelcontextprotocol/server-filesystem ~/projects
claude mcp add github -- npx -y @modelcontextprotocol/server-github
claude mcp add --scope project postgres -- npx -y @modelcontextprotocol/server-postgres $DATABASE_URL
claude mcp list
claude mcp logs github

Global vs project scope

The global scope (default) configures the MCP in ~/.claude/settings.json. The project scope adds it to .claude/settings.json in the current directory. Prefer project for sensitive access tokens.

`claude config` commands

Command	Description
`claude config list`	List all active configuration options
`claude config get <key>`	Show the value of an option
`claude config set <key> <value>`	Set a configuration option
`claude config add <key> <value>`	Add a value to a list
`claude config remove <key> <value>`	Remove a value from a list

# Examples
claude config set model claude-sonnet-4-5
claude config get model
claude config list

Next steps

CLI: Complete flag reference: All flags for the claude command
settings.json: Complete guide: Detailed configuration options
Environment variables: All recognized variables

Back

Overview

CLI reference