Tips & shortcuts

A few things that make Claude Code much nicer to use once you know them.

Keyboard

Shortcut	What it does
Shift+Tab	Cycle through permission modes (default → auto-accept → plan)
Esc then Esc	Edit your previous message
Ctrl+R	Toggle verbose tool output
!command	Run a shell command in the current session
#text	Add text to project memory (CLAUDE.md)

Slash commands

Type / at the start of a message to see available commands. Useful built-ins:

• /help — list everything
• /clear — start fresh, drop conversation history
• /compact — summarize and continue (saves context window)
• /cost — see token usage and $ for the session
• /model — switch model mid-session

CLAUDE.md

Drop a CLAUDE.md at the root of your repo. Claude reads it automatically every session — use it for project conventions, build commands, gotchas. Short and specific beats long and exhaustive.

Permission modes

• Default — asks before every tool use. Safest.
• Auto-accept edits — file edits run without prompts; commands still prompt.
• Plan mode — read-only; Claude can explore but not change anything. Great for big refactors where you want a plan first.

Toggle with Shift+Tab.

Run agents in parallel

For independent research tasks (e.g. "find all callers of X" + "find all tests for Y"), ask Claude to spawn agents in parallel. It's much faster than serial.