MCP server that exposes Proxmox VM consoles via noVNC so agentic AI models can drive boot-to-desktop workflows through the Proxmox API.
Get up and running in minutes
git clone https://github.com/Andreansx/VibeConsole-MCP.git
cd VibeConsole-MCP
npm install
npm run build
cp .env.example .env
# Edit .env with your Proxmox credentials
Everything you need to control VMs through AI agents
Connect to VM consoles via noVNC through Proxmox API with full keyboard and mouse control
Send keyboard input and capture real JPEG/PNG screenshots for AI agent interaction
Add coordinate grids to screenshots for spatial awareness and grid-based clicking
Get text with precise bounding boxes, grid cell locations, and regex search support
Click at grid references (e.g., "K9") or find and click text via OCR automatically
Wait for boot stages, detect system state, and monitor VM lifecycle events
Start, stop, reboot VMs and manage snapshots directly from the agent
Proper RFB protocol parsing with Sharp library for crystal-clear screen captures
Configure for Claude Desktop, Cursor, or any MCP client
{
"mcpServers": {
"vibeconsole": {
"command": "node",
"args": ["/path/to/VibeConsole-MCP/dist/index.js"],
"env": {
"PROXMOX_HOST": "pve",
"PROXMOX_PORT": "8006",
"PROXMOX_TOKEN": "vibeconsole@pve!mcp-token",
"PROXMOX_SECRET": "your-token-secret",
"DEFAULT_NODE": "pve",
"DEFAULT_VMID": "108"
}
}
}
}