Pvetui CLI

pvetui CLI

pvetui is a terminal UI and CLI for Proxmox VE. When invoked with a subcommand it runs non-interactively, making it suitable for scripts and AI agent workflows.

Installation

# Via Go (recommended)
go install github.com/devnullvoid/pvetui/cmd/pvetui@latest

# Via skill (Claude Code / any skills.sh-compatible agent)
npx skills add devnullvoid/pvetui

When to Use

• Querying node status, resource usage, or uptime
• Listing, filtering, or inspecting VMs and LXC containers
• Starting, stopping, shutting down, or restarting guests
• Creating VMs and LXC containers
• Migrating guests between nodes
• Executing commands inside a running guest without needing SSH to it
• Opening interactive shells on nodes or inside guests
• Listing and managing storage content (ISOs, templates, backups, disk images)
• Downloading ISOs, appliance templates, or OCI images into storage
• Restoring guests from vzdump backups
• Searching, inspecting, and installing Proxmox Community Scripts when community-scripts is enabled
• Listing recent cluster task history
• Scripting or automating Proxmox operations without root access to a node

Prerequisites

• pvetui installed and on $PATH
• Config at ~/.config/pvetui/config.yml with at least one profile
• API token or password auth configured in the profile
• For guests exec / guests shell on LXC containers: ssh_user configured in the profile
• For guests shell on QEMU VMs: ssh_user (and optionally vm_ssh_user) configured
• For community-scripts install: plugins.enabled includes community-scripts, and node SSH settings are configured

Quick Reference

Global Flags

These flags work with every subcommand:

Output Format

Default (JSON unless configured): structured output to stdout; errors as JSON to stderr; non-zero exit on failure.

Table (--output table): aligned human-readable output to stdout.

Set cli.default_output: table in config or PVETUI_CLI_DEFAULT_OUTPUT=table to make table output the default. Prefer --output json explicitly when parsing output in scripts or agents, because users may configure table as their default.

Task-Producing Commands

Commands that trigger a Proxmox task (guests create, guests migrate, storage content delete, storage download *, storage restore) block until the task completes by default, then include status and exit_status in the output. Pass --no-wait to return the task UPID immediately without waiting.

{
  "vmid": 105,
  "node": "pve01",
  "upid": "UPID:pve01:...",
  "status": "complete",
  "exit_status": "OK"
}

A non-"OK" exit_status causes a non-zero process exit.

Community Scripts

The community-scripts command group is available only when the opt-in plugin is enabled:

plugins:
  enabled:
    - community-scripts

Use search before installing and show to inspect the upstream metadata, source repo, script path, and documentation links.

# Search by name, slug, or description
pvetui community-scripts search nextcloud
pvetui community-scripts search docker --output table

# Show one script by exact slug or name
pvetui community-scripts show nextcloud

# Preview and then deploy with validated var_* overrides
pvetui community-scripts plan grafana --node pve01 --yes \
  --set var_hostname=grafana --set var_cpu=2 --set var_ram=2048 \
  --set var_container_storage=local-lvm --set var_template_storage=local

pvetui community-scripts deploy grafana --node pve01 --yes \
  --set var_hostname=grafana --set var_brg=vmbr0 --set var_net=dhcp \
  --set var_cpu=2 --set var_ram=2048 --set var_disk=8 \
  --set var_ssh=yes --set 'var_tags=monitoring;grafana' \
  --set var_container_storage=local-lvm --set var_template_storage=local

# Install a tools/addon script inside an existing LXC
pvetui community-scripts plan dockge --guest 200 --output table
pvetui community-scripts install dockge --guest 200

install/deploy SSHes to the selected node and runs the same Community Scripts installer flow as the TUI. It resolves SSH settings from the node source profile, active profile, or global ssh_user; set --ssh-user when needed. Use --node for scripts that create resources from, or modify, a Proxmox node. Use --guest <vmid> for tools/addon scripts such as Dockge, Coolify, and Komodo that should run inside an existing LXC; pvetui resolves the guest's node and runs the script through pct exec. Upstream installer output is streamed to stderr so stdout can contain the final JSON/table result. search/show output includes a derived target value: node-create, node, node-or-guest, or unknown.

For agent-driven deployments, prefer plan first, then deploy --yes with explicit --set var_*=value overrides. For unattended LXC deploys, include var_container_storage and var_template_storage; otherwise upstream scripts can open a storage picker and fail without a TTY. var_container_storage selects the LXC rootfs storage, var_template_storage selects where the OS template is downloaded, var_disk is the LXC rootfs size in GB, var_brg selects the bridge, var_net=dhcp requests DHCP, and var_vlan sets the Proxmox net0 VLAN tag. When both storage overrides are present, pvetui temporarily seeds /usr/local/community-scripts/default.vars on the target node so upstream first-run defaults do not prompt; the file is restored if it existed and removed if pvetui created it. --yes disables TTY allocation, selects the upstream default preset, skips the Community Scripts host-update prompt, uses sudo -n for non-root SSH users, and feeds empty stdin so unexpected prompts fail instead of hanging. Supported overrides are the Community Scripts allowlisted variables, including var_hostname, var_cpu, var_ram, var_disk, var_brg, var_net, var_gateway, var_vlan, var_container_storage, var_template_storage, var_ssh, var_tags, var_unprivileged, var_nesting, var_fuse, var_tun, and related var_* settings.

If the profile uses a non-root ssh_user, unattended Community Scripts installs require passwordless sudo for the escalation command. pvetui runs sudo -n su - root -c ... when --yes is used. On each Proxmox node, add a sudoers drop-in like this, adjusting the username and su path from command -v su:

Cmnd_Alias PVETUI_COMMUNITY_SCRIPTS = /usr/bin/su - root -c *, /bin/su - root -c *
youruser ALL=(root) NOPASSWD: PVETUI_COMMUNITY_SCRIPTS

This grants broad root command execution via su - root -c, which Community Scripts installation effectively requires. Prefer a dedicated automation account and SSH key, or connect as root.

Nodes

# List all nodes (JSON)
pvetui nodes list

# List nodes in human-readable table
pvetui nodes list --output table

# Show a specific node
pvetui nodes show pve01
pvetui nodes show pve01 --output table

# Open an interactive SSH shell on a node
pvetui nodes shell pve01

JSON shape — nodes list:

[
  {
    "name": "pve01",
    "ip": "10.0.0.10",
    "online": true,
    "cpu_usage": 0.12,
    "memory_used": 7.238216400146484,
    "memory_total": 31.174354553222656,
    "uptime": 864000,
    "version": "8.2.2",
    "kernel_version": "6.8.4-2-pve",
    "source_profile": "prod"
  }
]

cpu_usage is a fraction (0.0–1.0). Memory values are in GiB. uptime is seconds. source_profile is populated in aggregate group mode.

Guests (VMs and Containers)

# List all guests
pvetui guests list

# Filter by node, status, or type
pvetui guests list --node pve01
pvetui guests list --status running
pvetui guests list --type qemu
pvetui guests list --node pve01 --status running --type lxc

# Emergency/degraded mode: query only one node instead of full cluster guest inventory
pvetui guests list --node pve01 --node-local --status running

# Show a specific guest by VMID
pvetui guests show 100

JSON shape — guests list:

[
  {
    "id": 100,
    "name": "web-server",
    "node": "pve01",
    "type": "qemu",
    "status": "running",
    "ip": "10.0.0.100",
    "template": false,
    "tags": "prod;web",
    "source_profile": "prod"
  }
]

type is "qemu" or "lxc". status is "running", "stopped", or "paused". ip may be empty if the guest agent is not running. template: true guests cannot be started/stopped.

Lifecycle Operations

pvetui guests start 100       # start a stopped guest
pvetui guests shutdown 100    # graceful ACPI shutdown (preferred)
pvetui guests stop 100        # force power-off (data loss risk)
pvetui guests restart 100     # graceful restart

# Direct targeting skips guest discovery when you already know the node/type
pvetui guests start 100 --node pve01 --type qemu
pvetui guests shutdown 200 --node pve01 --type lxc

JSON shape — lifecycle:

{
  "vmid": 100,
  "operation": "shutdown",
  "upid": "UPID:pve01:00001234:...",
  "node": "pve01"
}

The upid is a Proxmox task ID. Use pvetui tasks list to monitor task completion.

Use --node <node> --type qemu|lxc on start, stop, shutdown, or restart when full guest discovery is slow or degraded but the guest's current node and type are known.

Guest Delete

Permanently deletes a guest and all its associated disks. The guest must be stopped first unless --force is passed.

# Delete a stopped guest (waits for completion)
pvetui guests delete 108

# Also remove the VMID from backup and replication job configs
pvetui guests delete 108 --purge

# Force-delete a running guest (data loss risk)
pvetui guests delete 108 --force

# Return the task UPID immediately without waiting
pvetui guests delete 108 --no-wait

JSON shape — delete:

{
  "vmid": 108,
  "node": "pve01",
  "upid": "UPID:pve01:...",
  "status": "complete",
  "exit_status": "OK"
}

Guest Create

# Create a VM (VMID auto-assigned if omitted)
pvetui guests create vm \
  --node pve01 \
  --name myvm \
  --disk-storage local-zfs \
  --disk-size 32

# With full options
pvetui guests create vm \
  --node pve01 \
  --name myvm \
  --vmid 105 \
  --memory 4096 \
  --cores 4 \
  --disk-storage local-zfs \
  --disk-size 32 \
  --iso local:iso/debian-12.iso \
  --bridge vmbr0 \
  --start

# Create an LXC container (package name resolved to latest template automatically)
pvetui guests create lxc \
  --node pve01 \
  --hostname myct \
  --rootfs-storage local-zfs \
  --template debian-12-standard

# LXC with full options
pvetui guests create lxc \
  --node pve01 \
  --hostname myct \
  --rootfs-storage local-zfs \
  --template debian-12-standard \
  --memory 1024 \
  --swap 512 \
  --cores 2 \
  --rootfs-size 16 \
  --bridge vmbr0 \
  --start

# Return the task UPID immediately without waiting
pvetui guests create vm --node pve01 --name myvm --disk-storage local-zfs --disk-size 32 --no-wait

Key flags — guests create vm:

Key flags — guests create lxc:

JSON shape — guests create:

{
  "vmid": 105,
  "name": "myvm",
  "node": "pve01",
  "type": "qemu",
  "upid": "UPID:pve01:...",
  "status": "complete",
  "exit_status": "OK"
}

Guest Migrate

# Migrate a guest (mode selected automatically)
pvetui guests migrate 100 pve02

# Force online/offline mode for QEMU (not valid for LXC)
pvetui guests migrate 100 pve02 --online
pvetui guests migrate 100 pve02 --offline

# Move migrated disks or LXC rootfs volumes to a target storage when supported
pvetui guests migrate 100 pve02 --offline --target-storage shared-ssd
pvetui guests migrate 200 pve02 --target-storage shared-ssd

# Wait longer for large disk copies
pvetui guests migrate 100 pve02 --offline --target-storage shared-ssd --wait-timeout 2h

# Return task UPID immediately
pvetui guests migrate 100 pve02 --no-wait

Migration mode is selected automatically:

• QEMU running → online migration
• QEMU stopped → offline migration
• LXC (any state) → restart migration

--target-storage is valid for QEMU offline migrations and LXC restart migrations. Proxmox may keep volumes on storage it considers shared; use pvetui tasks list and guests show or node tooling to verify final placement. --wait-timeout controls how long pvetui waits for the migration task before returning a timeout error; it does not cancel the underlying Proxmox task.

JSON shape — guests migrate:

{
  "vmid": 100,
  "name": "web-server",
  "source_node": "pve01",
  "target_node": "pve02",
  "mode": "online",
  "upid": "UPID:pve01:...",
  "status": "complete",
  "exit_status": "OK"
}

Guest Exec

Execute a command inside a running guest without SSH access to the guest itself.

# QEMU VM — uses the QEMU guest agent (API only, no node SSH needed)
pvetui guests exec 100 "uptime"
pvetui guests exec 100 "df -h"
pvetui guests exec 100 "systemctl status nginx"

# LXC container — uses pct exec over SSH to the node (ssh_user must be configured)
pvetui guests exec 200 "uptime"
pvetui guests exec 200 "cat /etc/os-release"

# Custom timeout (default 30s)
pvetui guests exec 100 "apt-get update" --timeout 120s

# Table output (useful for human review)
pvetui guests exec 100 "df -h" --output table

Requirements:

• QEMU: guest must be running with the QEMU guest agent enabled and responding.
• LXC: guest must be running; ssh_user must be configured in the profile; the agent user needs permission to run pct exec (or be root).

OS detection (QEMU only):

• Windows guests (ostype starting with win): command is wrapped in powershell.exe -NoProfile -NonInteractive -ExecutionPolicy Bypass -Command <cmd>
• All other guests: command is wrapped in /bin/sh -c <cmd>
• LXC always uses /bin/sh -c <cmd>

JSON shape — exec:

{
  "vmid": 100,
  "command": "uptime",
  "stdout": " 10:23:01 up 10 days,  2:34,  0 users,  load average: 0.10, 0.08, 0.05\n",
  "stderr": "",
  "exit_code": 0,
  "duration_ms": 412
}

A non-zero exit_code means the command ran but failed. A transport/SSH error causes a non-zero process exit with a JSON error on stderr instead.

Guest and Node Shell

Opens an interactive terminal session. These commands take over stdin/stdout and are not suitable for scripting — use exec for non-interactive commands in agents.

pvetui nodes shell pve01          # SSH to a Proxmox node
pvetui guests shell 100           # shell inside a guest (LXC: pct enter; QEMU: SSH to VM IP)

Tasks

# Last 20 tasks (default)
pvetui tasks list

# Last 50 tasks
pvetui tasks list --recent 50

# Human-readable table
pvetui tasks list --output table

JSON shape — tasks list:

[
  {
    "upid": "UPID:pve01:00001234:...",
    "node": "pve01",
    "type": "qmstart",
    "status": "OK",
    "user": "root@pam",
    "starttime": 1710000000,
    "endtime": 1710000005,
    "source_profile": "prod"
  }
]

starttime and endtime are Unix timestamps. status is "OK" on success or an error string on failure. A task with no endtime is still running.

Storage

# List all storages (one row per node/storage pair)
pvetui storage list
pvetui storage list --node pve01
pvetui storage list --output table

# Show a specific storage
pvetui storage show pve01 local-zfs

JSON shape — storage list:

[
  {
    "name": "local-zfs",
    "node": "pve01",
    "type": "zfspool",
    "content": "images,rootdir",
    "used": 53687091200,
    "total": 214748364800,
    "active": true
  }
]

Storage Content

# List all content in a storage
pvetui storage content list pve01 local

# Filter by content type: iso, vztmpl, backup, snippets, images
pvetui storage content list pve01 local --type iso
pvetui storage content list pve01 local --type backup

# Delete a content item (explicit volid required)
pvetui storage content delete pve01 local local:iso/old-debian.iso
pvetui storage content delete pve01 local local:iso/old-debian.iso --no-wait

JSON shape — storage content list:

[
  {
    "volid": "local:iso/debian-12.iso",
    "name": "debian-12.iso",
    "type": "iso",
    "size": 658505728,
    "ctime": 1700000000,
    "vmid": 0
  }
]

Storage Download

# Download an ISO from a URL (content type inferred from extension)
pvetui storage download url pve01 local https://example.com/debian-12.iso

# Override filename or content type
pvetui storage download url pve01 local https://example.com/image.bin \
  --filename my-image.iso \
  --content-type iso

# Download an appliance template by package name (latest version resolved automatically)
pvetui storage download template pve01 local debian-12-standard

# Or use a full filename to pin a specific version
pvetui storage download template pve01 local debian-12-standard_12.7-1_amd64.tar.zst

# Narrow resolution by section when multiple templates match
pvetui storage download template pve01 local myapp --section turnkeylinux

# Pull an OCI image
pvetui storage download oci pve01 local registry.example.com/myimage:latest

# Any download can return UPID immediately
pvetui storage download url pve01 local https://example.com/debian-12.iso --no-wait

Content type inference from URL extension:

JSON shape — storage download:

{
  "node": "pve01",
  "storage": "local",
  "url": "https://example.com/debian-12.iso",
  "filename": "debian-12.iso",
  "upid": "UPID:pve01:...",
  "status": "complete",
  "exit_status": "OK"
}

Storage Restore

Restore a guest from a vzdump backup. Destructive — overwrites the target VMID's config and disks.

# Dry-run: print what would be restored, exit without making changes
pvetui storage restore pve01 local local:backup/vzdump-qemu-100-2024.tar.zst 100

# Actually restore (requires --confirm)
pvetui storage restore pve01 local local:backup/vzdump-qemu-100-2024.tar.zst 100 --confirm

# Override inferred guest type (normally inferred from volid prefix)
pvetui storage restore pve01 local local:backup/unknown.tar.zst 101 --confirm --type lxc

# Return UPID immediately
pvetui storage restore pve01 local local:backup/vzdump-qemu-100.tar.zst 100 --confirm --no-wait

Guest type is inferred from the volid prefix: vzdump-qemu-* → qemu, vzdump-lxc-* → lxc. Use --type to override when inference fails.

JSON shape — storage restore:

{
  "node": "pve01",
  "storage": "local",
  "volid": "local:backup/vzdump-qemu-100-2024.tar.zst",
  "vmid": 100,
  "type": "qemu",
  "upid": "UPID:pve01:...",
  "status": "complete",
  "exit_status": "OK"
}

Profiles and Multi-Cluster

All subcommands respect --profile. The profile can be a single connection profile or an aggregate group name.

# Use the "prod" profile
pvetui --profile prod guests list

# Use the "all-clusters" aggregate group (fans out across all member nodes)
pvetui --profile all-clusters guests list --status running

# List available profiles
pvetui --list-profiles

In aggregate group mode, source_profile in JSON output identifies which cluster each resource belongs to. All reads fan out concurrently; write operations (lifecycle, exec, create, migrate) are routed to the correct cluster automatically.

Error Handling

Errors are written as JSON to stderr and the process exits non-zero:

{"error": "guest 999 not found"}

Check exit code to detect failure; parse stderr JSON for the message. Stdout is always valid JSON (or empty) on success.

# Example: check exit code in shell
if ! pvetui guests show 100 > /dev/null 2>&1; then
  echo "Guest not found"
fi

# Example: capture error message
err=$(pvetui guests show 999 2>&1 >/dev/null)

Common Agent Patterns

# Find all running VMs on a specific node
pvetui guests list --node pve01 --status running --type qemu

# Emergency drain inventory that avoids full cluster guest discovery
pvetui guests list --node pve01 --node-local --status running --output json

# Get IP addresses of all running guests
pvetui guests list --status running | jq '.[].ip'

# Check if a specific guest is running
pvetui guests show 100 | jq -r '.status'

# Create an LXC and wait for it to finish, then get its VMID
result=$(pvetui guests create lxc --node pve01 --hostname myct --rootfs-storage local-zfs --template debian-12-standard)
vmid=$(echo "$result" | jq -r '.vmid')

# Migrate a guest and verify it landed on the target
pvetui guests migrate 100 pve02
pvetui guests show 100 | jq -r '.node'   # should now be "pve02"

# Emergency node drain pattern
pvetui guests list --node pve01 --node-local --status running --output table
pvetui guests migrate 100 pve02 --offline --target-storage shared-ssd --wait-timeout 2h
pvetui guests shutdown 200 --node pve01 --type lxc
pvetui guests shutdown 101 --node pve01 --type qemu
pvetui tasks list --recent 20 --output table

# Download a template and create a container from it in sequence
pvetui storage download template pve01 local debian-12-standard
pvetui guests create lxc --node pve01 --hostname newct --rootfs-storage local-zfs --template debian-12-standard --start

# List all ISO files available on a node
pvetui storage content list pve01 local --type iso | jq '.[].volid'

# Find all backup files for a specific VM
pvetui storage content list pve01 local --type backup | jq '[.[] | select(.volid | contains("vzdump-qemu-100-"))]'

# Run a health check across multiple guests
for vmid in 100 101 102; do
  echo "=== $vmid ==="
  pvetui guests exec $vmid "systemctl is-active nginx"
done

# Find guests with a specific tag
pvetui guests list | jq '[.[] | select(.tags | contains("prod"))]'

# Get total memory across all nodes
pvetui nodes list | jq '[.[].memory_total] | add'

# Check storage usage across the cluster
pvetui storage list | jq '[.[] | {name, node, used_pct: (.used / .total * 100 | round)}]'

Troubleshooting