Every account starts with a 7-day trial. How to open it

Latte
Guide

How to approve Claude Code prompts from your phone

Your agent is waiting on “Can I run this?” and you're not at your desk. The options for answering from your phone, compared.

By Harsh Soni, maker of Latte · Updated

Claude Code asks before it runs commands or edits files outside what you've allowed. That's a good safety habit, but it means a long task can sit on one prompt for an hour while you're out. Here are the ways to answer from your phone.

Option 1: Claude Code Remote Control

Anthropic's Remote Control lets you pick up a Claude Code session that's running on your Mac from the Claude app or the web, on plans that include it. You see the conversation, answer prompts and keep typing.

  • Good: official, full conversation view, nothing extra to install.
  • Watch out: it's Claude Code only, and the session still runs on your Mac, so the Mac has to stay awake. Close the lid and Remote Control loses it.

Option 2: A notification hook (alerts only)

Claude Code supports hooks: commands it runs at points in its lifecycle. A Notification hook can ping your phone when Claude needs you, for example through ntfy. Add this to ~/.claude/settings.json:

{
  "hooks": {
    "Notification": [
      { "hooks": [{ "type": "command", "command": "curl -s -d 'Claude Code needs you' ntfy.sh/<your-private-topic>" }] }
    ]
  }
}

You'll know when it's waiting, but you still have to get back to the Mac (or SSH in) to answer. Use a long, random topic name: anyone who knows it can read the messages.

Option 3: SSH and tmux

Run Claude Code inside tmux, turn on Remote Login in System Settings, and attach from an SSH app on your phone over a VPN like Tailscale. It's the most flexible option and works for any agent, but typing in a terminal on a phone is fiddly, and you need the Mac awake and reachable.

Option 4: Latte

Latte (which I make) uses Claude Code's hooks to send permission prompts and questions to your own Telegram bot or ntfy topic, with buttons:

  • Approve or deny, seeing what the agent wants to run first.
  • Answer multiple-choice questions, or type your own answer.
  • Prompts only go to your phone when you're away. At your desk, the normal prompt shows in the terminal.
  • It also keeps the Mac awake with the lid closed, so the session is still there when you answer.
  • Messages go from your Mac straight to Telegram or ntfy, not through Latte's server.

Phone approvals work with Claude Code today. For Codex and Grok, Latte sends status and “finished” alerts.

Compared

Answer promptsWorks lid closedAgents
Remote ControlYes, full chatNeeds the Mac awakeClaude Code
Notification hookNo, alert onlyNeeds the Mac awakeClaude Code
SSH + tmuxYes, in a terminalNeeds the Mac awakeAny
LatteYes, with buttonsYesClaude Code (alerts for Codex, Grok)
Tip

These aren't either-or. You can use Remote Control for the conversation and Latte to keep the Mac awake with the lid closed. See keeping Claude Code running with the lid closed.

Common questions

Can I approve Claude Code permission prompts from my phone?
Yes. Claude Code's Remote Control lets you answer from the Claude app on plans that include it, and Latte sends prompts to your own Telegram bot or ntfy topic with Approve and Deny buttons.
Does Remote Control work when my MacBook lid is closed?
Only if the Mac stays awake. The session runs on your Mac, so when the lid closes and the Mac sleeps, Remote Control cannot reach it.
Is it safe to approve commands from a phone?
It is as safe as approving at your desk if you can see what the agent wants to run before you tap. Deny anything you can't read in full, and keep risky commands off your allow list.
Close the lid. Keep shipping.

Latte keeps Claude Code, Codex and Grok working with your MacBook closed, sends approvals to your phone, and lets the Mac sleep the moment the work is done.

Start your 7-day trialNo card needed · $2 a month after