How to answer Claude Code or Codex from your phone

You start an agent on the computer at home and leave. An hour later your phone shows that it stopped to ask for permission to run a command. It stays stopped until you answer, and you're nowhere near the desk.

Each of the big agents has a way to answer from your phone. They just don't work the same way:

• Claude Code: run /remote-control in the session, then open it in the Claude app or at claude.ai/code. It needs a Pro, Max, Ultra, Team, or Enterprise plan.
• Codex: set up Control this Mac or PC in the ChatGPT desktop app, pair your phone with the QR code, and answer from the ChatGPT mobile app.
• Any agent: turn on SSH on the computer, reach it from outside your network, and keep the agent in tmux. A terminal app on the phone does the rest.
• The catch: the computer has to stay awake and online. If it sleeps, every one of these stops.
• The other catch: each agent has its own app and its own setup. With three agents, that's three.

How do I answer Claude Code from my phone?

Use Remote Control, Claude Code's own feature. In a session you've already started, type:

/remote-control

(/rc is the short form.) The session keeps running on your computer, with your files, MCP servers, and settings. Your phone only shows it and sends messages. Open the session by name in the Claude app under Code, or at claude.ai/code. A session that's online shows a computer icon with a green dot.

You can also start it from the shell:

claude remote-control      # server mode, shows a URL; press space for a QR code
claude --remote-control    # a normal interactive session with Remote Control on

To be told when it needs you, open /config and turn on Push when actions required. Install the Claude app and sign in with the same account.

What to know before you rely on it:

• It needs a Pro, Max, Ultra, Team, or Enterprise plan with a claude.ai login. API keys aren't supported.
• Outside server mode, each claude process supports one remote session, and that process has to keep running. Close the terminal and the session ends.
• A session you started without it can't be reached until you run /remote-control in it, unless you turn on Enable Remote Control for all sessions in /config.

The full list is in Anthropic's Remote Control documentation.

How do I answer Codex from my phone?

Codex does it through the ChatGPT apps, and setup starts on the computer, not in the terminal. You can't set it up from the Codex CLI or the IDE extension.

1. In the ChatGPT desktop app on the computer, open Settings → Connections → Control this Mac or PC, choose Set up or Add, and approve remote access.
2. Scan the QR code with your phone. It opens the ChatGPT app.
3. Confirm the same account and workspace. The computer then appears under Codex in the app, or Remote if your app still uses that name.

From the phone you can continue chats, answer questions, approve commands, and read diffs and terminal output. The work still runs on the computer, with its files and tools.

The computer must be awake, online, and signed in. On a Mac laptop, remote access can stay available with the lid open and power connected; with the lid closed it also needs an external display, and choosing Sleep ends it. Each phone is paired with each computer it controls. The rest is in OpenAI's remote connections documentation.

Can I use SSH and tmux for any agent?

Yes, and it's the one method that doesn't care which agent you run. You connect to the computer, then attach to a terminal where the agent is already running. It takes more setup than the two above.

On the Mac, turn on SSH: open System Settings → General → Sharing, then turn on Remote Login. Under Allow access for, choose Only these users and add yourself.

From outside your home network you also need a way to reach the Mac. Don't open its SSH port to the internet. Use a VPN or a mesh network such as Tailscale, so only your own devices can connect.

Then run the agent inside tmux, which keeps the terminal alive when you disconnect:

tmux new -s agents     # start a session named "agents", then run claude or codex inside it
# Ctrl-b, then d       # detach: the agent keeps running
tmux attach -t agents  # from your phone's SSH app, pick it up again

Anthropic's own documentation suggests the same for Claude Code over SSH: use tmux or screen so the session survives a dropped connection.

This works, but it's the roughest on a phone. You're typing into a full terminal with a touch keyboard, and the keys an agent needs, Esc, Tab, Ctrl, and the arrows, aren't on it. You also set up the SSH app, the network, and the Mac yourself.

Why does my agent stop answering when I'm away?

Almost always, the computer went to sleep. Locking the screen doesn't stop an agent; idle sleep does, and so does closing the lid. Remote Control for Claude Code and Codex, and SSH, can only reach a computer that's awake and online.

None of the agents keep the Mac awake between turns, and that's exactly when you're waiting to answer from the phone. How to keep your Mac awake for Claude Code and Codex covers it step by step.

What if I run Claude Code, Codex, and Grok together?

Then there's no single place to answer. Claude Code's remote lives in the Claude app, Codex's in the ChatGPT app, and a Grok session on your own computer has no remote from xAI that I could find in its documentation. Over SSH you can reach all three, but you have to find the right tmux session yourself and read the terminal to see which one is waiting.

With one agent, any of the methods above is fine. With three, the problem is no longer how to connect, it's which one is waiting for you.

Can PonyMux answer all of them from one place?

Yes. PonyMux runs the real CLI of each agent in its own Terminal, so Remote Control doesn't depend on any agent offering its own remote. It's free on macOS and Windows.

1. In the sidebar, select the cloud next to Agents, then Sign In…, and sign in with GitHub or Google.
2. Scan the QR code on the card with your phone. my.ponymux.com opens with your agents already listed.
3. Every agent running in PonyMux shows up in one list with the same status lights: Claude Code, Codex, or Grok. Open the one that's waiting.
4. Opening an agent only shows its Terminal. Choose Start control to send keys to it. On a phone, a key bar adds the esc, tab, ctrl, and arrow keys the keyboard lacks.

One device controls an agent at a time. On the computer, a Stop button takes control back whenever you want.

What it doesn't do: it doesn't wake a sleeping computer, so Keep Awake is worth turning on before you leave. There's no end-to-end encryption, and no approval cards that answer for you; you read the Terminal and type the answer. The server relays the screen and keystrokes and saves none of them. On the Free plan, one computer shares its agents at a time.

For setup, see Set up Remote Control. Remote Control on your phone covers the key bar and notifications, and How Remote Control stays safe lists exactly what is and isn't protected.

Download PonyMux, free for macOS and Windows: ponymux.com/download