← All tutorialsGET STARTED

How to monitor Claude Code and Codex from your phone

Connect your machine with PeerBand and add its MCP URL to ChatGPT or Claude. On your phone, ask the app "What are my agents doing?" For alerts, add a Slack or Discord webhook that fires when an agent is blocked or idle.

Your options

There are a few ways to keep an eye on agents away from your desk:

  • The agent's own remote feature, such as Claude Code Remote Control or Codex in the ChatGPT app. These work well for one tool's sessions.
  • SSH from your phone into tmux. It works with anything, but it's awkward on a small screen and doesn't tell you when something happens.
  • PeerBand: every agent in Herdr, of any kind and on any of your machines, through the ChatGPT or Claude app you already have, plus alerts when one needs you.

Set it up once, at your desk

npm install -g peerband
peerband setup

Then add https://mcp.peerband.com/mcp as a custom connector in ChatGPT or Claude on the web. The connector is then available in the mobile app wherever your plan supports connectors.

Check in from your phone

  • "Which agents are blocked?" lists agents waiting on a permission prompt or a question.
  • "What did the agent in the api workspace print last?" shows a bounded excerpt of recent output.
  • "Tell it yes, go ahead" sends a message to that exact agent, with your approval.

The PeerBand dashboard also works on a phone. It shows every machine and agent live, with no refresh needed.

Get pinged instead of checking

Add a webhook in the dashboard for the blocked status and you get a Slack or Discord message the moment an agent waits for you. See the Slack tutorial.

Questions

Does my computer need to stay on?

Yes. The agents run on your machine, so it must be awake with Herdr running. The connector reconnects automatically after sleep or network changes.

Can I approve a permission prompt from my phone?

You can send the agent a message, such as an answer to its question. Whether a message answers an approval dialog depends on the agent's interface; read its output first to see what it's waiting for.

Open dashboard Read the docs