Take over your VPS

Run the installer

The same command as any other install:

curl -fsSL https://deplo.build/install.sh | bash

It finds the other panel and asks:

[ !] Dokploy is already running on this machine.
     Deplo can only install here by taking its place: it brings your projects
     across, then removes Dokploy. Two panels cannot share 80 and 443.

  -> Migrate off Dokploy and let Deplo replace it? [y/N]

Answer no and nothing is installed. Answer yes and Deplo installs on a free port, leaving the other panel exactly where it is.

--yes does not answer this one: replacing somebody's platform is not a default. To install without being asked, set DEPLO_TAKEOVER=dokploy (or coolify).

Open the address it prints

The installer prints the way in and exits. The terminal is yours again: the half that needs root on the host - moving the ports, restarting Docker, removing the old panel - is a systemd unit it leaves behind, deplo-takeover.service, which waits for the dashboard to say so.

Next
  1  Open http://deplo-cb00710b.nip.io/setup?key=... and create your account.
     No SSH needed - it goes in through Dokploy's own proxy on port 80.
  2  Bring your projects over from Dokploy, or start clean.
     The dashboard takes the machine when you say so, and opens by itself after.

That address is Deplo's address, before and after: your domain if you gave one, the generated deplo-<hex>.nip.io otherwise. Until the cutover it answers over plain http through the old panel's own proxy - Deplo writes one route into it, which it reloads by itself - and after the cutover the same host answers over https from Deplo's proxy. Nothing to re-open, nothing to re-type.

To watch the host side, journalctl -u deplo-takeover -f; its transcript is /var/log/deplo-takeover.log. Re-running /opt/deplo/install.sh only puts the unit back if it is missing, and prints the address again.

Create your account, and choose

After the account and the team, Deplo opens the takeover wizard and nothing else: until the ports are its own, there is nothing to deploy onto.

Its first step asks the one question that cannot be undone later:

• Bring your data over - your apps, databases and their data move to Deplo, and the old panel comes off the machine once they are here. This is the rest of this page.
• Start clean - the old panel comes off the machine with everything it ran: apps, teams, networks, images and its directory. Nothing comes across. The volumes its apps held stay on the disk until you remove them yourself, the same as after a migration. Deplo skips straight to Take over.

The wizard's steps are Choose, Connect, Install, Review, People, Take over and Done. A step opens only when the one before it has finished, so you can never land on a step with nothing in it.

Migrate

The panel it is replacing is already filled in. Paste its API token, pick what you want to bring across, and start.

The migration runs in Review, and stays there: the progress, the log, and then the report of what came across. If anything needs a person, the report says so and you acknowledge it before the wizard moves on.

Take over the machine

Take over opens when there is something to take over for: a migration that finished, or the Start clean choice. Until then the step is closed - there is no button to press and nothing to explain.

With several teams on the list it stays shut until the last one is over, and then asks you to tick I have no other teams on this panel. Nothing here can start the old panel again once its ports are gone, so that tick is the last chance to go back for a team you forgot.

A service whose data copy failed gets a second tick of its own, naming it: the old panel still holds the only copy of that data. Its volume stays on the disk after the takeover, but nothing in Deplo reads it. Go back to Review and copy the data again instead of ticking it, unless you mean to lose it. The takeover also waits while any migration is still moving.

One confirmation does the whole cutover, and you type the panel's name for it. On the host, deplo-takeover.service:

• stops the old panel and the apps it ran
• copies its certificates into its own store
• moves Traefik onto 80 and 443, with the dashboard served behind it
• takes the old panel off the disk: its containers, the workloads it ran and their networks, its images and its directory. The volumes those workloads held stay on the disk, so a copy that went wrong can still be recovered, and so do the backups the panel wrote to its own disk - the installer prints where they are
• restarts Docker, which applies the address pools the install wrote and puts Docker's own DNS back after the old panel's swarm is left

The apps go down because two copies of one app cannot both hold a port, and the ones the old platform ran are holding exactly the ports Deplo's proxy is about to take. Deplo's own container is not touched: only who answers on 80 and 443 changes.

The wait stays in the step, and the page follows the dashboard by itself: onto the home, at the https address, once the old panel is gone and the address answers. It does not move earlier on purpose - the removal restarts Docker, and a page that had already moved would die with it. There is nothing to open and nothing to reload.

If the port move fails, the unit puts the ports back and starts the old panel again, and the step says why, in the installer's words, next to Try again. The reason names what is holding the port: a container of yours that grabbed 80 while the proxy was down is the usual one. Remove it, then Try again.

Done

The browser lands on the dashboard's home, on the same address over https, with a note that the old panel is gone. Ports 80 and 443 are Deplo's now, so a real certificate issues within seconds of the move; the page waits for it before it goes there, so there is no warning to click through.

deplo-takeover.service removes itself once the machine is Deplo's.

Deploy your apps

Deplo never deploys anything for you. Your apps arrived with their configuration and their data, and you deploy them one at a time, checking each one.