Approve Claude Code without touching the terminal
Claude asks for permission at the worst moments — while you're in the browser, the editor, or another project. Here's how to answer from the top of your screen instead of hunting for the right tab.
Why this is harder than it sounds
Claude Code’s permission prompt lives in the terminal, and its hooks offer no remote-answer API for it. Anything that claims to approve from outside is really doing one of two things: intercepting tool calls before the prompt exists, or answering the real prompt in the terminal on your behalf.
Interception pings you for everything
A PreToolUse hook fires for every tool call — including ones your permission settings would auto-approve. Route those to a popup and an autonomous session pings you dozens of times an hour about scripts it was always allowed to run.
The quiet way: listen, then answer the real prompt
The Notification hook fires only when Claude actually asks. NapCat listens there: a genuine ask takes over the notch as one big card, and Allow or Deny jumps to the session’s exact terminal tab and answers the real prompt — the same keys you’d press yourself. Ignore the card and the terminal behaves exactly as it always did; nothing is ever answered without your click.
NapCat does all of this from your Mac’s notch — for five dollars, once.
Download — free 2-day trial