Run Codex on an always-on developer pod

A cloud computer of your own, on your own private network. Codex keeps working with your laptop shut, and you pick its sessions up from your laptop or your phone.

Get your developer pod$24/mo, cancel anytime · 4 vCPU, 8 GB, 80 GB

Or let Codex do it: paste this into it.

Get me an Everpod developer pod and walk me through setting it up, following https://everpod.ai/docs/api#developer-pod

Codex gets a computer of its own.

Not a laptop you have to leave open: a whole Ubuntu machine you administer, always on, set up and secured before you log in.

Your repos, your installs and your files stay where you left them, for ever.

Tick Claude Code too, and both come installed. Ubuntu, with you as its administrator. Docker, git, Node, Python, Chrome and tmux are there, and you install anything else as you would on a laptop.

Three ways to keep Codex running.

Your laptopstops when the lid shuts
A VPSsetting it up and securing it is yours
A developer pod$24/mo, arrives set up and private

No laptop left open, no server to secure. Your pod arrives set up, and stays on.

Codex CLI on a Linux server, step by step: the guide

OpenAI’s hosted tasks, or a computer of your own

OpenAI’s hosted tasks

✓work in a workspace prepared from your GitHub repositories
✓keep working while your computer is asleep
your local files, the processes you leave running and your private network stay behind

A developer pod

A developer pod is one machine that stays yours: what you installed, the servers you left running, Docker and your work in progress are where you left them tomorrow.

✓always on, on your own private network
✓your tools and your servers beside Codex

OpenAI’s documentation on remote access advises: “add an always-on computer or SSH host when you need continuous access.” A developer pod is that SSH host, and it never sleeps. The ChatGPT desktop app on a Mac or PC can open projects on it too.

Codex runs here as it does on your laptop.

Sign in with ChatGPT, with no browser on the machine

  • ✓Approve the one-time code in any browser, on your laptop or your phone
  • ✓It needs one switch on ChatGPT’s side: device code sign-in, under Settings, Security and login, or the workspace permission an admin turns on
  • ✓An OpenAI API key works too, billed by use
alex@atlas:~$ codex login --device-auth

Welcome to Codex [v0.160.0]
OpenAI's command-line coding agent

Follow these steps to sign in with ChatGPT using device code authorization:

1. Open this link in your browser and sign in to your account
   https://auth.openai.com/codex/device

2. Enter this one-time code (expires in 15 minutes)
   INHF-PYASB

Its sandbox works from the first command

  • ✓The commands Codex chooses write only to your project’s folder
  • ✓On a fresh Ubuntu 24.04 server the sandbox fails until bubblewrap may create user namespaces; a developer pod comes with that set
alex@atlas:~/app$ codex sandbox -c sandbox_mode=workspace-write -- touch notes.txt /etc/hello
touch: cannot touch '/etc/hello': Read-only file system

Which ChatGPT plans include it

  • ✓Codex in the terminal comes with ChatGPT Plus and the plans above it
  • ✓Free and Go have it only in the desktop app
PlusProBusinessEnterpriseEduFree and Go: desktop app only

All you do is name your machine.

We build it, install Codex and secure it. About 10 minutes later, you log in.

1

Name it and choose your agents

Name the machine, choose your username on it, and tick the agents you want installed.

2

Confirm your email and pay

$24 a month, cancel anytime. We start building the moment your payment goes through.

3

Join your network and log in

About 10 minutes later it is built, and we email you. Join it to your Tailscale network from its page, switch off its key expiry, and log in.

Private by default

Your pod arrives on your own private network, closed to the internet, and yours alone.

On your private network only

Reached only from your own devices, over Tailscale. No open ports and no public SSH: a server your agent starts is reachable from your devices and from nowhere else.

Yours alone

A whole private computer, one customer per machine. Nothing on it is shared with other customers.

Only you have a login

Once you have finished connecting the machine, which ends with your first login, we remove the access we used to set it up. From then on Everpod has no login on it and no key.

Backed up every day

The whole disk is backed up every day, and the last seven days are kept. We restore one when you ask.

Security updates install themselves

The machine is yours to administer, and its security updates install on their own.

Deletion is permanent

When you cancel, the machine runs to the end of the paid month, is switched off, and is permanently deleted 7 days later, backups included.

You don’t do any of this. Your pod arrives on your private network, closed to the internet.

What your developer pod costs.

$24/month
✓an always-on cloud computer of your own, with admin rights
✓Claude Code, Codex or both installed, with Docker, git, Node, Python and Chrome
✓reached over your own Tailscale network, with nothing open to the internet
✓no Everpod login on it once you have connected it
✓4 vCPU, 8 GB of memory, 80 GB of disk
✓daily automatic backups
✓cancel anytime
You sign in to each agent with your own subscription or API key. No model usage is included, and your sign-ins never pass through us.
Get your developer pod

Before you decide.

Is model usage included?
No. Codex runs on your own ChatGPT plan or an OpenAI API key, billed by OpenAI, and the price is for the machine.
Who else can log in to it?
Only you. Once you have finished connecting the machine, which ends with your first login, we remove the access we used to set it up. From then on Everpod has no login on it and no key.
Can I steer it from the ChatGPT app on my phone?
Through a Mac or PC, yes. The phone app connects to a Mac or Windows computer running the ChatGPT desktop app, which can work on the pod as an SSH host, so that computer has to stay awake. To reach the pod with nothing else switched on, use Tailscale’s app and any SSH app on your phone. Codex remote control goes through both routes.
Does it work with a ChatGPT Business workspace?
Yes, once device code sign-in is allowed for the workspace, which is a permission its admin turns on. If the sign-in stops with a message about it, the device code guide walks through it.
Can I run Claude Code on it too?
Yes. Tick both when you start it, or install Claude Code later as on any computer. Claude Code on a developer pod says how it runs there.
What are its specs?
4 vCPU, 8 GB of memory, 80 GB of disk, with room for about 8 coding agents working at once. One size is on offer today; the developer pod’s page asks which bigger size you would want.
What happens if I cancel?
Cancel anytime from your account. The machine runs to the end of the paid month, is switched off, and is permanently deleted 7 days later, backups included.

Codex, on a computer that stays on.

$24/mo, cancel anytime · 4 vCPU, 8 GB, 80 GB