Claude Remote Approver

claude-remote-approver

Approve or deny Claude Code permission prompts from your phone.

Problem

Claude Code asks for permission before running tools like Bash, Write, and Edit. These prompts require you to be sitting at your terminal. If you step away, Claude Code stalls until you come back and press "y".

claude-remote-approver sends each permission prompt as a push notification to your phone via ntfy.sh. You tap Approve or Deny, and Claude Code continues immediately -- no terminal required.

How it works

1. Claude Code invokes the hook, piping the tool request as JSON to stdin.
2. cli.mjs hook sends a notification to your ntfy topic with Approve, Always Approve, and Deny action buttons.
3. The hook subscribes to a response topic (-response) via server-sent events.
4. When you tap a button on your phone, ntfy.sh publishes your decision to the response topic.
5. The hook reads the decision and writes {"behavior":"allow"} or {"behavior":"deny"} to stdout. If the notification fails or times out, the hook returns {"behavior":"ask"} so Claude Code falls back to the CLI prompt.
6. Claude Code proceeds accordingly.

AskUserQuestion support

When Claude Code calls the AskUserQuestion tool, the hook sends the question to your phone as a notification. Each option appears as an action button you can tap. If the question has more than 3 options, they are split across multiple notifications. If no response is received before the timeout, the hook falls back to the CLI prompt so you can answer at your terminal.

Always Approve

When Claude Code sends a permission_suggestions field with the hook request (indicating the tool can be auto-approved in future), the notification shows three buttons: Approve, Always Approve, and Deny.

- Approve -- Allow this one request.
- Always…