Questions

Why not Claude Code on the web?

The web version fits an isolated task on one repository. It does not give you a machine: no database next to the code, no headless browser, no service running overnight, nothing that persists once the session ends. Pupitre gives you the machine, and your subscription works there as it is.

How is it different from Coolify or Dokploy?

They deploy applications to production. Pupitre sets up the machine your coding agents develop on. It gives them a server of their own, on all the time, where they build, run tests, open headless browsers and work in parallel worktrees, next to the databases and services of your projects. That load leaves your laptop. The projects under development get an address you can open from anywhere: on your computer through the app’s SSH session, or at a route of your domain through exposure.cloudflare or exposure.caddy, or on your tailnet through exposure.tailscale. See projects.

Pupitre does no deployment pipeline, no zero-downtime release and no scaling. The two fit together: develop on a Pupitre server, and ship with the tool you already use for production.

Can I run production on it?

Nothing stops you: the server is yours, and anything outside the catalogue installs the usual way. But Pupitre is built as a development machine, where agents work unattended as dev, and it makes no promise about a production workload. A separate server for production is the safer layout.

Do I have to work inside the app?

No. The server is an ordinary Ubuntu machine reached over SSH as dev. The Claude desktop app takes an SSH connection with the key Pupitre generated, the ChatGPT desktop app reads the host aliases of your ~/.ssh/config to run Codex there, and ssh works the way it always has. The app stays the place for installing, configuring and watching the machine. See your own client.

Can I use my existing server?

Yes, if it runs Ubuntu 22.04 or 24.04. The probe reports what is already installed and the modules only manage what they installed themselves. A machine that already has a database set up by hand keeps it; Pupitre will not adopt it, and will not fight it either.

Does it work on Windows?

The desktop app runs on macOS, Windows 11 and Ubuntu desktop. The server is always Ubuntu — Pupitre installs nothing on a Windows VPS.

Do I need an account?

Yes. The app asks you to sign in at its first launch, and it enrols a server for an organisation: free up to 3 servers, with a licence beyond. The account holds your organisation, its servers and its members. See install the app and licence.

Does it cost anything?

Not for the first 3 servers of an organisation: no card, no trial, no subscription. Beyond them, a licence is granted on request. You pay your host for the server and your model provider for your agents, as before. See licence.

What happens if I lose my laptop?

Revoke the device in the console. Its key is removed from every server it was on at the agent’s next check-in. Your other devices keep working; nothing else has to be rotated. The agent never removes the last key of a server: if that laptop was the only one opening it, take it back from your host’s console, see your key.

Can two people share a server?

A server is assigned to one person. Two people who need the same machine share it the way they always have, over SSH, with their own keys added by hand. Pupitre’s model is one machine per developer, because that is what makes revocation clean.

Where is my data?

On your server. When an agent works, what it reads goes from the server to the model provider you chose, on your account, never through Pupitre. The platform stores your account, your organisation and the public keys of your devices; of each server, its name, address, architecture and agent version, the services and terminal sessions it runs, its health readings over seven days, and the status of its backups. It stores no code, no secrets, no logs, and no contents from your machines.

Who can see a project I publish?

Nobody without an access key. A published address asks for one. The app adds your computer’s key when it opens the address, and you create others for a colleague, a client or a script. A process can stay public to receive webhooks. See access to addresses.

Can I read the agent’s source?

Yes. The code of the app, the agent and the platform is public, under the Apache 2.0 licence with the Commons Clause: you may read it, change it, build the agent yourself and run your own platform, but not sell Pupitre or a service built mainly on it. See the licence. What the agent does is also documented here, module by module, and everything it installs is a standard tool you can inspect on your own machine. The security page lists what the agent can and cannot do, with the command that checks each claim.

What happens if Pupitre shuts down?

Your server keeps running. Without the platform, the app and the agent keep working for seven days. After that, the agent stops accepting changes and leaves everything that runs in place. SSH, your editors, your services, databases, projects and backups need nothing from us, and sudo pupitred uninstall removes the agent. See security and uninstall.