Proxmox MCP

An MCP (Model Context Protocol) server for managing Proxmox VE infrastructure through AI agents. Control your virtual machines, containers, storage, and snapshots via natural language.

WARNING: This project is under active development and is NOT intended for production use. This MCP server can perform destructive operations on your Proxmox infrastructure, including deleting virtual machines and containers. Data loss is possible. Use at your own risk. Always test in a non-production environment first.

---

Features

Discovery

• list_nodes — Cluster node overview
• list_vms / list_lxc_containers — Instance inventory per node
• list_storage / list_storage_content — Storage pools, ISOs, and templates
• get_instance_config — Full guest configuration
• get_lxc_interfaces — Container network interfaces
• get_resource_usage — Performance metrics (CPU, memory, network, disk)

Management

• power_control — Start, stop, shutdown, reboot VMs and containers
• create_lxc / create_vm — Provision new infrastructure
• execute_lxc_command — Run shell commands inside containers
• set_vm_cloudinit — Configure cloud-init (user, SSH keys, network)

Protection

• create_snapshot / rollback_snapshot / list_snapshots — Snapshot management
• delete_instance — Permanently delete a VM or container (disabled by default, requires double confirmation)
• get_task_status / get_task_log — Monitor background operations
• get_mcp_logs — Server-side log inspection

---

Quick Start

# 1. Clone and set up
git clone https://github.com/josky-com/proxmox-mcp.git
cd proxmox-mcp
./setup.sh

# 2. Configure credentials
cp .env.example .env
nano .env    # Set your Proxmox URL, token name, and token value

# 3. Test connectivity
python scripts/test_connection.py

Environment Variables

Variable	Description	Default
PROXMOX_URL	Proxmox API base URL (e.g., https://192.168.1.100:8006/api2/json)	—
PROXMOX_TOKEN_NAME	API token ID (e.g., root@pam!mcp-token)	—
PROXMOX_TOKEN_VALUE	API token secret	—
PROXMOX_VERIFY_SSL	Verify SSL certificates	true
PROXMOX_ALLOW_DANGER	Enable critical/destructive tools like delete_instance	false

---

AI Client Configuration

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "proxmox": {
      "command": "/absolute/path/to/proxmox-mcp/.venv/bin/python3",
      "args": ["-m", "proxmox_mcp"]
    }
  }
}

Claude Code

Add to your project's .mcp.json:

{
  "mcpServers": {
    "proxmox": {
      "command": "/absolute/path/to/proxmox-mcp/.venv/bin/python3",
      "args": ["-m", "proxmox_mcp"]
    }
  }
}

---

Security

Safety Tiers

All tools are classified into safety tiers:

Tier	Behavior	Examples
Safe	Executes without confirmation	list_nodes, list_vms, get_instance_config
Restricted	Requires explicit user confirmation (confirmed=true)	power_control, create_lxc, create_vm, execute_lxc_command
Critical	Blocked unless PROXMOX_ALLOW_DANGER=true + confirmation	delete_instance

Contract Validation

Every Proxmox API response is validated against Pydantic schemas. If the Proxmox API returns unexpected data (e.g., after a version upgrade), the server returns an API_CONTRACT_DIVERGENCE error instead of silently passing bad data.

Input Sanitization

All identifiers (node names, storage names, snapshot names) are validated against strict patterns. VMIDs must be positive integers. Shell commands are parsed through shlex to prevent injection.

---

Project Structure

proxmox-mcp/
├── src/proxmox_mcp/
│   ├── __main__.py           # Entry point (python -m proxmox_mcp)
│   ├── server.py             # FastMCP instance and tool registration
│   ├── config.py             # Environment variables, client init
│   ├── models.py             # Pydantic response schemas
│   ├── api/
│   │   ├── client.py         # Async HTTP client for Proxmox API
│   │   └── permissions.py    # Permission mapping for 403 diagnostics
│   ├── core/
│   │   ├── logger.py         # Rotating file logger with redaction
│   │   ├── safety.py         # Safety policy enforcement
│   │   └── sanitization.py   # Input validation
│   └── tools/
│       ├── __init__.py       # safety_checked decorator, shared helpers
│       ├── discovery.py      # 10 tools: nodes, VMs, containers, storage, config, logs
│       ├── lifecycle.py      # 4 tools: power, create LXC/VM, delete
│       ├── snapshots.py      # 3 tools: list, create, rollback
│       ├── cloudinit.py      # 1 tool: cloud-init configuration
│       ├── exec.py           # 1 tool: execute commands in containers
│       └── metrics.py        # 1 tool: resource usage metrics
├── docs/
│   ├── api_manifest.md       # API contract registry (20 tools)
│   ├── architecture.md       # System design overview
│   └── requirements.md       # Functional requirements
├── scripts/
│   ├── test_connection.py    # Manual connectivity testing
│   └── admin_dashboard.py    # Live infrastructure overview
├── tests/                    # pytest suite (sanitization, safety)
├── pyproject.toml            # Build config and dependencies
├── setup.sh                  # Automated setup script
└── .env.example              # Credential template

---

Tech Stack

• Python 3.10+
• MCP SDK (mcp) — FastMCP server framework
• Pydantic — Response validation and contract enforcement
• httpx — Async HTTP client
• Transport — stdio (JSON-RPC)

---

Trademark Notice

"Proxmox" and the Proxmox logo are registered trademarks of Proxmox Server Solutions GmbH (Vienna, Austria). This project is not affiliated with, endorsed by, or sponsored by Proxmox Server Solutions GmbH. This is an independent, third-party tool that integrates with the Proxmox VE API.

---

License

This project is licensed under the MIT License.