You also find this cheat sheet attached!
---
Claude Code has a long list of supported CLI commands & flags, shortcuts and settings.
In my experience, the defaults are fine for most cases, except for permissions (e.g., prevent access to .env or similar files) or sandboxing - hence my more detailed coverage in the past videos.
But here's a short list of commands and settings you should be aware of. Many of them were or will be covered in the course lectures, too.
CLI Commands & Flags
For all available commands and flags, see the official docs.
claude / claude "my prompt" => Start a new Claude Code session, optionally with an initial prompt.
claude -p "my prompt" => Query Claude Code with "my prompt" and quit session once it's done (no ongoing session).
claude -c => Continue most recent session.
claude --agent DocsExplorer => Start a new session with a custom agent (instead of the default one). Custom agents will be covered later in the course.
claude --allowedTools "Read" "Write" => Skip permission confirmation dialog for specified permissions.
claude --disallowedTools "Write" => Disallow certain tools (specified tools are removed from Claude Code's context => it doesn't know about them).
claude --dangerously-skip-permissions => Skip ALL permissions confirmation dialogs. Use with great caution!
claude --append-system-prompt "Always check if redundant code can be deleted" => Append an instruction to the default system prompt. Can be used as session-specific alternative to instructions placed in CLAUDE.MD (CLAUDE.MD will be covered later).
claude --model opus => Sets the default model for the current session. Can be changed in the session via /model, or generally via settings (see below). For valid options, see the official list.
claude --permission-mode plan => Start in plan mode (default is "default" mode where Claude Code asks for permissions). For available options, see the official list. "Plan" mode will be covered later.
claude --remote "Add dark mode" => Start a remote (web) session (will be covered later in the course).
claude --system-prompt "You are a React.js expert" => Replace the entire default system prompt with a custom one.
Claude Code Interaction
When inside a Claude Code session, you can use the following shortcuts for additional control. For all available shortcuts and commands, see the official docs.
SHIFT + ENTER / OPTION + ENTER / CTRL + J => Enter new line. Exact command depends on platform and terminal software used.
SHIFT + TAB => Switch through different modes ("default", "write permissions", "plan").
CTRL + C => Cancel current input or generation.
ESC => Cancel current generation (can be used to inject a new prompt into an ongoing task).
ESC + ESC => Restore the code prior to the last action performed by Claude Code.
OPTION + P / ALT + P=> Switch model. Alternative to /model which can be useful if you already entered a prompt.
ARROW keys (left / right / top / down) => Cycle through options or questions (left / right) or past message (up / down).
CTRL + O => Toggle verbose output (more / less details).
CTRL + B => Move task to background. Important: Claude Code also automatically moves tasks (e.g., exploration tasks) to the background and it's typically quite smart about that.
CTRL + V / CMD + V / ALT + V=> Insert text or image into prompt. The exact shortcut depends on your platform and terminal.
CTRL + O => Toggle verbose output (more / less details).
Crucial Commands
/help => List available commands and get usage help.
/model => Choose the active model for the current session.
/clear => Clear session context window (i.e., start a new session, effectively).
/compact => Compact (summarize) current session context history and clear it thereafter.
/config => Open interactive settings menu.
/context => View statistics about current context window & context usage.
/usage => View current usage for your active plan (i.e., how much usage remains until the next reset). For more details about plans and usage, see the official help pages.
/init => Analyze project and create initial CLAUDE.MD file (covered later).
/mcp => View and manage installed MCP servers.
/permissions => View and update / change permissions.
/rewind => Rewind (undo) to earlier point in conversation, same as ESC + ESC.
/statusline => Configure the Claude Code status line.
/teleport => Resume a remote Claude Code session (covered later).
Settings
You can configure (project-specific) settings also interactively via /settings. Or in your global, project-specific or local (non-Git-controlled) settings.json files (see past lectures).
For all available settings, see the official docs.
{ "permissions": {...} } => Manage permissions for all sessions. See previous lectures and official permissions settings docs.
{ "model": "opus" } => Manage default AI model used for new sessions.
{ "alwaysThinkingEnabled": true } => Whether to enabled "advanced thinking" mode or not. Is on by default. Turning it off will likely yield worse results.
{ "hooks": { } } => Manage hooks (covered later in the course).
{ "env": { "IS_DEMO": 1 } } => Environment variables that will be applied to every session.