ProxmoxMCP-Plus

Execute a bounded shell command directly on a Proxmox physical node via SSH. This tool is disabled unless the selected target explicitly sets ssh.allow_node_commands=true. Requires configured SSH credentials, a node present in the authenticated Proxmox node inventory, and command_policy approval. Host command allow_patterns must match the entire command, not just a prefix or substring. Use narrowly scoped expressions; permissive wildcards may still allow unintended shell operations. It is classified as a high-risk operation and respects target and client grants. Parameters: node* - Exact Pro…

Execute commands in a VM via QEMU guest agent. Parameters: node* - Host node name (e.g. 'pve1') vmid* - VM ID number (e.g. '100') command* - Shell command to run (e.g. 'uname -a') Example: {"success": true, "output": "Linux vm1 5.4.0", "exit_code": 0}

Read public cluster join metadata and certificate fingerprints. Only documented public fields are returned. Authentication credentials and arbitrary extensions are omitted. Proxmox errors are reported.

Read recent cluster-wide log entries. Parameters: max_entries - Maximum number of entries to return (default: 50, max: 1000) Each entry includes: time, node, user, tag (service), pri (priority), msg. Example: get_cluster_log max_entries=100

Get overall Proxmox cluster health and configuration status. Example: {"name": "proxmox", "quorum": "ok", "nodes": 3, "ha_status": "active"}

Get the full configuration of an LXC container. Returns network interfaces, mounts, features, CPU/memory limits, startup options and more. Parameters: node* - Proxmox node name (e.g. 'pve') vmid* - Container ID (e.g. '101') Example: {"vmid": "101", "hostname": "valkey", "cores": 1, "memory": 1024, "net0": "name=eth0,..."}

Get the current IP address(es) of a running LXC container. Queries /nodes/{node}/lxc/{vmid}/interfaces - works with DHCP (no static IP needed). Parameters: node* - Proxmox node name (e.g. 'pve') vmid* - Container ID (e.g. '101') Returns: {"vmid": "101", "name": "valkey", "interfaces": [...], "primary_ip": "10.1.0.101"}

List LXC containers across the cluster (or filter by node). Parameters: - node (optional): Node name to filter (e.g. 'pve1') - include_stats (bool, default false): Fetch per-container live CPU/memory stats - include_raw (bool, default false): Include raw Proxmox API payloads for debugging - format_style ('pretty'|'json', default 'pretty'): Pretty text or raw JSON list Notes: - Live stats from /nodes/{node}/lxc/{vmid}/status/current. - If maxmem is 0 (unlimited), memory limit falls back to /config.memory (MiB). - If live returns zeros, the most recent RRD sample is used as a fallback. - Field…

Read the firewall log of a VM (qemu) or container (lxc). Parameters: node* - Node hosting the guest (e.g. 'pve') vmid* - VM/container ID (e.g. 100) vm_type - Guest type: 'qemu' (default) or 'lxc' limit - Maximum number of log lines to return (default: 100, max: 1000) start - Start line for pagination (0-based, optional) since - Show entries since this UNIX epoch timestamp (optional) until - Show entries until this UNIX epoch timestamp (optional) Example: get_guest_firewall_log node='pve' vmid=100 vm_type='qemu' limit=50

Get the current state of a tracked job by job_id. Parameters: job_id* - Stable job identifier returned by long-running tools refresh - Poll Proxmox before returning the job state (default: false)

Get the next free VM/container ID in the cluster. Uses GET /cluster/nextid. Call it right before clone_vm or create_vm so the target ID is one Proxmox itself considers unused. Example: {"vmid": "105"}

Read CPU, I/O wait, memory, storage and latest observed RRD disk/network metrics for one node. Missing optional metrics produce partial results. Does not execute shell commands or list processes.

Read the host firewall log of a Proxmox node. Parameters: node* - Node name (e.g. 'pve', 'pve1') limit - Maximum number of log lines to return (default: 100, max: 1000) start - Start line for pagination (0-based, optional) since - Show entries since this UNIX epoch timestamp (optional) until - Show entries until this UNIX epoch timestamp (optional) Example: get_node_firewall_log node='pve' limit=50

Get detailed status information for a specific Proxmox node. Parameters: node* - Name/ID of node to query (e.g. 'pve1') Example: {"cpu": {"usage": 0.15}, "memory": {"used": "8GB", "total": "32GB"}}

Read storage capacity and active/rootdir status on one specified node. Optionally filter by storage ID. Available bytes do not guarantee migration capacity, especially on thin-provisioned storage.

Read syslog entries from a Proxmox node. Parameters: node* - Node name (e.g. 'pve', 'pve1') limit - Maximum number of log lines to return (default: 100, max: 1000) start - Start line for pagination (0-based, optional) since - Show entries from this date/time onward (format: YYYY-MM-DD or YYYY-MM-DD HH:MM or YYYY-MM-DD HH:MM:SS, optional) until - Show entries up to this date/time (same format as since, optional) service - Filter by service name (e.g. 'pvedaemon', 'pveproxy', optional) Example: get_node_syslog node='pve' limit=50 service='pvedaemon'

Read node epoch time, local time and timezone to diagnose PVE ticket clock skew. Read-only; requires node and accepts an optional configured target.

List all nodes in the Proxmox cluster with their status, CPU, memory, and role information. Example: {"node": "pve1", "status": "online", "cpu_usage": 0.15, "memory": {"used": "8GB", "total": "32GB"}}

List storage pools across the cluster with their usage and configuration. Example: {"storage": "local-lvm", "type": "lvm", "used": "500GB", "total": "1TB"}

Get the log output of a specific Proxmox task by its UPID. Parameters: node* - Node name that ran the task (e.g. 'pve') upid* - Unique Process ID of the task (e.g. the Task ID returned by VM/container operations or list_jobs) start - Start line for pagination (0-based, optional) limit - Maximum number of log lines to return (default: 50, max: 500) Example: get_task_log node='pve' upid='UPID:pve:00001234:...' limit=100

Get the full configuration of a QEMU virtual machine. Returns hardware configuration including CPU, memory, disk, network, BIOS type, boot order, and more. This is equivalent to get_container_config but for QEMU VMs. Parameters: node* - Host node name (e.g. 'pve') vmid* - VM ID number (e.g. '100') Example: {"vmid": "100", "name": "ubuntu", "cores": 2, "memory": 4096, "scsi0": "local-lvm:vm-100-disk-0,size=20G"}

Get a running VM's network interfaces and IP addresses via the QEMU guest agent. Uses GET /nodes/{node}/qemu/{vmid}/agent/network-get-interfaces. The VM must be running with qemu-guest-agent active. Loopback is skipped; primary_ip is the first non-loopback IPv4 address. Parameters: node* - Host node name (e.g. 'pve') vmid* - VM ID number (e.g. '105') Example: {"vmid": "105", "primary_ip": "10.0.0.57", "interfaces": [{"name": "eth0", "mac": "bc:24:11:..", "ipv4": ["10.0.0.57"], "ipv6": ["fe80::..."]}]}

List all virtual machines across the cluster with their status and resource usage. Example: {"vmid": "100", "name": "ubuntu", "status": "running", "cpu": 2, "memory": 4096}

Import an existing qcow2/raw/vmdk image from Proxmox import-enabled storage into an unused VM SCSI slot via the asynchronous QEMU API. Requires an observed stopped VM, explicit confirmation and configured high-risk approval. Does not fetch URLs, overwrite disks or change boot order. Returns a submitted task, never completion. Inspect SUBMISSION_UNCERTAIN outcomes before any retry.

List available backups across the cluster. Parameters: node - Filter by node (optional) storage - Filter by storage pool (optional) vmid - Filter by VM/container ID (optional) Returns backups with timestamp, size, compression, and notes. Use the returned Volume ID with restore_backup.

List available network bridges (Linux, OVS and permitted SDN VNets) on a node for VM/LXC networking. Returns the Proxmox network records as JSON.

List available ISO images across the cluster. Parameters: node - Filter by node (optional) storage - Filter by storage pool (optional) Returns list of ISOs with filename, size, and storage location.

List tracked long-running jobs created by MCP tools. Parameters: status - Filter by job status (optional) tool_name - Filter by originating tool (optional) limit - Maximum rows to return (default: 100)

List all snapshots for a VM or container. Parameters: node* - Host node name (e.g. 'pve') vmid* - VM or container ID (e.g. '100') vm_type - Type: 'qemu' for VMs, 'lxc' for containers (default: 'qemu') Example: list_snapshots node='pve' vmid='100' vm_type='qemu'

List configured Proxmox targets without exposing credentials. A target name is required for other tools when multiple targets are configured.

List available OS templates for container creation. Parameters: node - Filter by node (optional) storage - Filter by storage pool (optional) Returns list of templates (vztmpl) with name, size, and storage. Use the returned Volume ID with create_container's ostemplate parameter.

Submit an LXC migration after fresh preflight checks. Requires confirm_migration=True; restart=True also requires confirm_downtime=True. Follows the configured target's high-risk approval policy. Returns a submitted task UPID and optional job ID, never completion. Never automatically retries; SUBMISSION_UNCERTAIN requires inspecting Proxmox tasks and guest location before any retry.

High-risk OFFLINE QEMU migration with local disks after read-only preflight and explicit confirmations of backup, workload drain, CPU compatibility and migration. The VM must already be stopped. Returns a submitted UPID/job ID, never completion. Never automatically retry uncertain submissions.

Poll the backing Proxmox task for a tracked job and refresh status/progress.

Read-only LXC migration assessment: checks nodes, rootfs storage and space, runtime, extra mounts/devices, HA and matching backups. Missing evidence blocks readiness. This advisory does not reserve capacity, test restoreability or migrate. Running guests require allow_downtime=True.

Read-only QEMU offline migration assessment using native Proxmox migration preconditions plus VM inventory, storage capacity, bridges and HA checks. Missing evidence blocks readiness. Requires VM.Migrate permission even for the native read. Backup restorability, workload drain and CPU compatibility need external verification. Does not move or stop a VM or reserve capacity.

Execute sandboxed async code using call_tool(name, arguments).

Return exact input schemas for discovered Proxmox tools.

Search Proxmox MCP tools by name or description.

Reconcile an interrupted retry after verifying its remote task or confirming that no task was submitted. Never submits a new operation.

Reset (restart) a virtual machine. Parameters: node* - Host node name (e.g. 'pve') vmid* - VM ID number (e.g. '101') Example: Reset VPN-Server with ID 101 on node pve

Restart LXC containers (reboot). selector: same grammar as start_container

Restore a VM or container from a backup. Parameters: node* - Target node for restore archive* - Backup volume ID (from list_backups output) vmid* - New VM/container ID for the restored machine storage - Target storage for disks (optional, uses original if not specified) unique - Generate unique MAC addresses (default: true) Example: restore_backup node='pve' archive='backup:backup/vzdump-qemu-100-2024_01_15.vma.zst' vmid='200'

Retry a tracked long-running job using its stored retry recipe. The same job_id is preserved and its attempt counter is incremented.

Rollback VM/container to a previous snapshot. WARNING: This will stop the VM/container and restore to the snapshot state! Parameters: node* - Host node name vmid* - VM or container ID snapname* - Snapshot name to restore vm_type - Type: 'qemu' or 'lxc' (default: 'qemu') Example: rollback_snapshot node='pve' vmid='100' snapname='before-update'

Set/replace the description (Notes field in the UI) of an LXC container. Uses PUT /nodes/{node}/lxc/{vmid}/config. Pass an empty string to clear the notes. Parameters: node* - Proxmox node name (e.g. 'pve') vmid* - Container ID (e.g. '101') description* - New notes text (replaces any existing notes) Example: set_container_description node='pve' vmid='101' description='GitLab Runner host (ct101-alpine)'

Set/replace the description (Notes field in the UI) of a QEMU VM. Uses PUT /nodes/{node}/qemu/{vmid}/config. Pass an empty string to clear the notes. Parameters: node* - Host node name (e.g. 'pve') vmid* - VM ID number (e.g. '100') description* - New notes text (replaces any existing notes) Example: set_vm_description node='pve' vmid='100' description='Decommissioned - see #123'

Shutdown a virtual machine gracefully. Parameters: node* - Host node name (e.g. 'pve') vmid* - VM ID number (e.g. '101') Example: Gracefully shutdown VPN-Server with ID 101 on node pve

Start one or more LXC containers. selector: '123' | 'pve1:123' | 'pve1/name' | 'name' | comma list Example: start_container selector='pve1:101,pve2/web'

Start a virtual machine. Parameters: node* - Host node name (e.g. 'pve') vmid* - VM ID number (e.g. '101') Example: Power on VPN-Server with ID 101 on node pve

Stop LXC containers. graceful=True uses shutdown; otherwise force stop. selector: same grammar as start_container timeout_seconds: 10 (default)

Stop a virtual machine (force stop). Parameters: node* - Host node name (e.g. 'pve') vmid* - VM ID number (e.g. '101') Example: Force stop VPN-Server with ID 101 on node pve

Update an existing LXC network interface: bridge, static IPv4/IPv6, DHCP and gateways. Preserves MAC, VLAN, firewall and other NIC settings. Use an empty gateway to remove it.

Update resources for one or more LXC containers. selector: same grammar as start_container cores: New CPU core count (optional) memory: New memory limit in MiB (optional) swap: New swap limit in MiB (optional) disk_gb: Additional disk size in GiB to add (optional) disk: Disk identifier to resize (default 'rootfs')

Inject or replace SSH authorized_keys for root in an LXC container. Uses pct exec via SSH to the Proxmox host - requires SSH to be configured. Parameters: node* - Proxmox node name (e.g. 'pve') vmid* - Container ID (e.g. '101') public_keys* - Newline-separated public key(s) to authorize mode - 'append' (default) or 'replace' Returns: {"success": true, "keys_added": 1}

Update sizing, cloud-init, ISO media, boot order and net0 bridge of an existing QEMU VM. iso_volume accepts a storage volume ID or none to eject. cdrom_device defaults to ide3; existing disks and cloud-init drives cannot be overwritten. boot_order uses semicolon-separated device names. network_bridge preserves the other net0 options. Uses PUT /nodes/{node}/qemu/{vmid}/config with only the supplied fields. This is the step between clone_vm and start_vm that makes a fresh clone reachable: inject SSH keys, set the cloud-init user, pick a static IP, resize memory. Parameters: node* - Host…