Hetzner and Proxmox VPS tickets, answered with live data

VPS tickets are short and urgent: "my server is down", "the VPS is very slow", "I lost the root password". The first useful reply needs facts that are not in WHMCS. Is the machine powered on? Has the provider locked it? Was it rebooted an hour ago? Is the CPU flat out? A technician opens the Hetzner Console or the Proxmox web interface to find out. Makli reads the same data through their APIs, so its first reply already contains it.

This post covers what Makli does with Hetzner Cloud servers and Proxmox VE virtual servers sold through WHMCS, and the limits it keeps.

The server is always the client's own

Makli only looks at servers that belong to the client on the ticket. It takes the service from that client's WHMCS account, never from a server ID typed into the conversation. If a client asks about a service that is not theirs, Makli refuses. If they have several VPS services, it asks which one.

For Hetzner, the link is the server ID recorded on the WHMCS service. If that field is empty, Makli can find the server in your Hetzner project by the service's dedicated IP. A server that matches only by its hostname is used for reading at most, never for a power or password action, because a client chooses their own hostname and a matching name proves nothing. Makli then tells your staff to fill in the missing server ID.

What Makli reads

Two small details matter more than they look. Hetzner reports CPU on a scale where 100 means one fully used vCPU, so Makli divides by the server's vCPU count before it says "the server is at 50%". And on Proxmox, a KVM guest's RAM figure can include the guest's file cache, so Makli says "check inside the OS" instead of declaring the server out of memory.

What Makli will not claim

Hypervisor data is useful, but it describes the virtual machine, not what runs inside it. Makli is written to keep that difference in every reply:

• "Running" is not "working". A powered-on server can still have a stopped web server or a full disk. Makli does not tell the client their server is fine from the power state alone.
• No samples is not low load. If Hetzner returns no metrics for the last hour, Makli says the load could not be judged.
• An empty history is not proof. Hetzner's action list shows only what was done through Hetzner. A crash or a reboot from inside the operating system never appears there.
• No hidden causes. If the Proxmox VM is running with normal usage and no failed tasks, the reply says the server itself is up and the problem is probably inside the OS or the application. It does not invent one.

What Makli can change on Hetzner

Proxmox support is read-only: Makli never starts, stops or changes a Proxmox VM. On Hetzner Cloud there are six actions that change something:

1. Power on a server that is off. Makli checks first; if it is already running, powering on would change nothing.
2. Reboot, a soft restart. The operating system has to respond to the signal, so Makli says the reboot was requested, not that the server is back.
3. Graceful shutdown, like pressing the power button once. Hetzner confirms only that the signal was sent, so Makli never reports the server as off until the power state says so.
4. Force power off, like pulling the plug. Unsaved data can be lost. This action starts Off, so Makli cannot use it until you turn it on.
5. Reset the root password. Hetzner can only do this on a running Linux server with the QEMU guest agent. Makli gives the new password to the verified account holder in the reply and keeps it out of notes and hand-offs. It also updates the password stored on the WHMCS service.
6. Turn backups on or off. Turning them on adds 20% to the server's Hetzner price, so Makli asks the client to accept the cost first. Turning them off deletes every existing backup of the server immediately, and Makli refuses to do it without an explicit confirmation from the client.

Who decides

Every change above starts on Ask me first, except force power off, which starts on Off. On a ticket, the action waits in your approval queue and no reply goes out until someone on your team decides. In live chat, by default nothing that changes the server runs; the client is offered a ticket instead. Power on and reboot are the only two that Makli may also do for a sender it has identified as the account holder but not fully verified. The password reset and the backup switch need a verified client. You can change each mode later, one action at a time. More on modes and approvals in how Makli keeps your WHMCS safe, and on why the reads come first in why a hosting support AI must check the server.

What you need to set up

• Hetzner Cloud: an API token for the project your client servers live in, and the Hetzner server ID on each WHMCS service.
• Proxmox VE: an API user or token that can read the VMs (VM.Audit), and the VM ID, node and type recorded for each VPS service. Proxmox shows an API user only its own tasks unless it also has Sys.Audit on the node, so Makli never treats an empty task list as proof that nothing failed.

Want to see what Makli would have said on your last ten VPS tickets? Test Makli runs on real tickets, reads only, and changes nothing. Start free and try it on yours.