That flag appears on power control, creation, deletion, snapshots, rollback, cloud-init and command execution, which makes an accidental call distinguishable from an executed one. delete_instance goes further, requiring a separate confirming VMID that has to match the target before anything is removed. Long operations return a UPID rather than a result, and the task status and task log tools are how a create that is still running is told apart from one that failed.
A Proxmox client covering cluster nodes, VMs and LXC containers, storage, snapshots, background tasks and resource metrics.
- Cluster nodes listed with the VMs and containers on each, their configuration, and an LXC's network interfaces.
- Storage per node, and the files held inside a store.
- LXC creation from a template with cores, memory, disk, network, hostname, password and SSH keys; QEMU VM creation with memory, network, disk and an onboot flag.
- Power control — start, stop, shutdown and reboot — and permanent deletion where the confirming VMID has to match the target.
- Snapshots listed, created with a name and description, and rolled back to.
- Cloud-init configuration for a QEMU VM: user, password, SSH keys and IP configuration.
- A shell command run inside an LXC container, background task status and logs, CPU, memory and network metrics over a timeframe, and the server's own logs.
A Proxmox deployment you can reach, and an account on it entitled to create, delete and power the instances you intend to touch.
One command plus a key — uvx proxmox-mcp, then supply credentials