HomeLab Monitor — Claude Code MCP server: what it does, how to install
Exposes read-only homelab telemetry to AI coding assistants, tracking Docker containers, multi-GPU loads, and hardware power costs.
claude mcp add --env HOMELAB_MONITOR_URL=<HOMELAB_MONITOR_URL> --env MCP_TRANSPORT=<MCP_TRANSPORT> --transport stdio homelab-monitor -- docker run -i --rm -e HOMELAB_MONITOR_URL -e MCP_TRANSPORT docker.io/sikamikaniko123/homelab-monitor:0.14.2What HomeLab Monitor does
This MCP server connects AI assistants to an existing HomeLab Monitor dashboard via a lightweight Docker container. It grants the model instant visibility into hosts, disks, and active containers queried across your machines over SSH. The tool reads metrics from NVIDIA, AMD, and Intel hardware, capturing clock speeds, thermal throttle causes from nvidia-smi, and individual fan RPMs.
Hardware power tracking converts raw wattage from RAPL and GPU drivers into actual monetary cost per process or container. For local machine learning workflows, it reports token throughput in Ollama and flags context sizes that spill from VRAM into system memory. Access is strictly read-only. That prevents the agent from modifying or rebooting any services. Running the container requires valid HOMELAB_MONITOR_URL and MCP_TRANSPORT environment values.
What it helps with
- inspect VRAM allocation and temperature across multiple GPUs
- calculate energy costs generated by specific containers or processes
- diagnose GPU throttling reasons during model training
- track token generation speeds and memory spillover in Ollama
- review host storage and uptime over SSH connections
Example prompts for the agent
- Show current fan speeds and temperatures across all GPUs in the lab.
- How much electricity did the database container consume today?
- Check if the local Ollama model fits entirely within VRAM.
How to install HomeLab Monitor
Platform CLI
claude mcp add --env HOMELAB_MONITOR_URL=<HOMELAB_MONITOR_URL> --env MCP_TRANSPORT=<MCP_TRANSPORT> --transport stdio homelab-monitor -- docker run -i --rm -e HOMELAB_MONITOR_URL -e MCP_TRANSPORT docker.io/sikamikaniko123/homelab-monitor:0.14.2Replace <…> with your own values
Platform CLI
codex mcp add homelab-monitor --env HOMELAB_MONITOR_URL=<HOMELAB_MONITOR_URL> --env MCP_TRANSPORT=<MCP_TRANSPORT> -- docker run -i --rm -e HOMELAB_MONITOR_URL -e MCP_TRANSPORT docker.io/sikamikaniko123/homelab-monitor:0.14.2Replace <…> with your own values
Config file
{
"mcpServers": {
"homelab-monitor": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"-e",
"HOMELAB_MONITOR_URL",
"-e",
"MCP_TRANSPORT",
"docker.io/sikamikaniko123/homelab-monitor:0.14.2"
],
"env": {
"HOMELAB_MONITOR_URL": "<HOMELAB_MONITOR_URL>",
"MCP_TRANSPORT": "<MCP_TRANSPORT>"
}
}
}
}Replace <…> with your own values
Config file
{
"servers": {
"homelab-monitor": {
"type": "stdio",
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"-e",
"HOMELAB_MONITOR_URL",
"-e",
"MCP_TRANSPORT",
"docker.io/sikamikaniko123/homelab-monitor:0.14.2"
],
"env": {
"HOMELAB_MONITOR_URL": "<HOMELAB_MONITOR_URL>",
"MCP_TRANSPORT": "<MCP_TRANSPORT>"
}
}
}
}Replace <…> with your own values
Platform CLI
gemini mcp add -e HOMELAB_MONITOR_URL=<HOMELAB_MONITOR_URL> -e MCP_TRANSPORT=<MCP_TRANSPORT> homelab-monitor docker run -i --rm -e HOMELAB_MONITOR_URL -e MCP_TRANSPORT docker.io/sikamikaniko123/homelab-monitor:0.14.2Replace <…> with your own values
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about HomeLab Monitor
How to install HomeLab Monitor in Claude Code?
Register the server by running `claude mcp add --env HOMELAB_MONITOR_URL=<HOMELAB_MONITOR_URL> --env MCP_TRANSPORT=<MCP_TRANSPORT> --transport stdio homelab-monitor -- docker run -i --rm -e HOMELAB_MONITOR_URL -e MCP_TRANSPORT docker.io/sikamikaniko123/homelab-monitor:0.14.2`. Replace the placeholders with your running instance address and transport method.
What does HomeLab Monitor do?
It gives your AI assistant read-only access to host telemetry, including Docker status, multi-GPU metrics, disk utilization, and energy costs.
Can the MCP server restart failing containers or run shell commands?
No, the server operates strictly in read-only mode. It supplies telemetry to the model without supporting any management or control operations.
