Your Laptop Does the Work. Your Phone Approves It.

A video render on my second Mac takes about fifteen minutes. For most of this year my answer to that was a small Telegram bot that pings my phone when a long job finishes, so I am not sitting at the desk watching a bar crawl across the screen. I still run it for other things.

For Claude Code, I did not need to build any of that. It has a feature called Remote Control, and it does something better than sending you a notification. It hands you the actual session on your phone.

That word “session” is worth pausing on, because everything here depends on it. A session is one ongoing conversation with Claude Code, running on your computer, with your files in front of it. Not a chat window in a browser. A running thing on your machine that can read your folders and do work. Remote Control lets you open that exact running thing on your phone, mid-task, and keep talking to it.

If you came out of the masterclass with Holt set up, this is what makes it portable. Setup takes about ninety seconds and no new software beyond the Claude app.

Where the work actually happens

There are two different things you can reach at claude.ai/code, and mixing them up causes most of the confusion.

One of them runs on Anthropic’s computers, somewhere in a data centre. The other, Remote Control, runs on your laptop. Your files stay on your laptop. Your settings stay on your laptop. Your Holt memory folder stays on your laptop. The phone is a window into work happening at your desk, not a second copy of it somewhere else.

Here is a small thing that makes it obvious. Type the @ symbol on your phone and it will start suggesting file names from the project folder on your laptop. Your phone does not have those files. Your laptop does, and it is the one answering.

This is exactly why it matters for Holt. Holt’s memory of you lives in a folder on your machine. Something running in a data centre cannot see that folder. Remote Control can see all of it, because it is the same session that has been reading that folder since you started it.

Getting connected

If Claude Code is already running and you want to carry on from the sofa, type this into it:

/remote-control

You can shorten it to /rc. It brings your current conversation along, so nothing is lost. It prints a link, and if you run the command a second time it shows you a QR code. Point your phone camera at that code and the session opens in the Claude app.

That is the whole setup. There is no account to create, no server to run, no port to open.

If you would rather it start out connected every time, type /config inside Claude Code and turn on the option called “Enable Remote Control for all sessions.” Then you never think about it again.

If you are using Holt

When you type holt, Holt gets your folder ready and then hands the screen over to a normal Claude Code session wearing a Holt name badge. Underneath, it is ordinary Claude Code. So /rc works inside Holt exactly the way it works anywhere else.

Start Holt in your project folder, type /rc, scan the code. Your memory, your skills, all of it, now reachable from a bus.

One thing to know. Remote Control is a Claude Code feature. Holt can also run on Codex or Gemini as its brain, and this does not apply to those. If you want Remote Control, keep Claude Code as your brain.

The step most people skip

Without notifications, Remote Control is a webpage you have to remember to check, which means you will not check it. With notifications, your phone taps you on the shoulder when the session needs you.

Four steps, once:

1. Install the Claude app on your phone.
2. Sign in with the same account you use on your laptop.
3. Say yes when your phone asks whether Claude can send notifications.
4. Back on your laptop, type /config in Claude Code and switch on both “Push when Claude decides” and “Push when actions required.”

The first one tells you when a long job finishes. The second tells you when Claude is stuck waiting for your permission to do something.

You can also just ask, in normal words, at the end of any instruction: “notify me when the tests finish.” And if you are sitting at the laptop typing, it will not send you a notification, because you are obviously already there.

What you can do from the phone

More than you would expect. You can type instructions. You can say yes or no when Claude asks permission to do something. You can send a photo from your camera roll, with or without a note about it. You can watch background helpers running and stop one if it has gone off track. You can see the list of changes Claude has made to your files but not yet saved permanently.

A few things stay on the laptop only, and they are the housekeeping ones you are unlikely to want from a phone anyway, like switching to an older conversation.

What quietly stops it working

Most people who cannot get this going hit one of these. In rough order of how often it happens.

You are on the free plan. Remote Control needs a paid Claude plan: Pro, Max, Team, or Enterprise. There is no way around this one.

You are signed in the wrong way. If you set Claude Code up with an API key rather than by logging into your Claude account, Remote Control will not start. Type claude auth login and sign in through your normal Claude account instead.

You closed the terminal. This is the big one, and it surprises people. The session lives on your laptop, so if you close the window it was running in, it stops, and your phone shows it as offline. Leave it open. Your laptop going to sleep is fine, though. Messages and permission requests wait in line and arrive when it wakes up.

You never approved the folder. The first time you run Claude Code in a project folder it asks whether you trust that folder. Answer it once, from inside an actual project folder rather than your home folder.

You are on a work laptop. Company setups often route Claude through their own systems or switch off the background checks Remote Control needs, and then it refuses to start. The error message will name the setting causing it, which you can pass to whoever manages your machine. On a Team or Enterprise plan there is also a switch an admin has to turn on before anyone can use this at all.

There is one more thing that is not a fault but you should know about it. While your phone is connected, the conversation itself is stored on Anthropic’s servers, because that is how the two screens stay in step and how it recovers when your wifi drops. Your files and the actual work stay on your machine. For most work that is a fine trade. For a folder full of a client’s confidential material, decide on purpose rather than find out later.

Where it changes the day

Nobody wants to rewrite a chunk of software with their thumbs, and working from your phone was never really the point.

The point is the moment Claude stops and asks for permission. You set something running, walk away, make tea, and it hits a step it wants you to approve. Before, that question sat on your screen for twenty minutes until you wandered back. Now it arrives on your phone, you tap yes, and the work carries on without you in the room.

The other point is catching things early. You glance at your phone, see it has gone in the wrong direction, and type two sentences to turn it around before it spends another ten minutes going further wrong. Fixing an agent early is cheap. Fixing it after an hour is not, and catching it early means being reachable, which is exactly what a laptop sitting on a desk in another room stops you from being.

Everything else is the same session you already had. It just stopped being tied to the chair.