ServerMind
Star on GitHub Install
Use case · Homelabs & self-hosting

A private AI copilot for your homelab

Docker stacks, Pi-hole, a media server, an Immich box, a NAS — your homelab does a lot, and you self-host precisely so your data stays yours. ServerMind fits that ethos: no cloud, no domain required, and it can run entirely offline on a local model.

The problem

You self-host to own your data — so a cloud "AI assistant" that phones home is a non-starter.

Homelabs sprawl. You've got containers you set up months ago, services across a couple of machines, and a mental map that's always slightly out of date. Checking on it means SSH-ing into each box and running the same handful of commands, and most monitoring tools are either heavyweight enterprise platforms or SaaS that wants your metrics shipped off-site.

What you actually want is something small, local, and private — that you can simply ask "is everything up?" without anything leaving the house.

How ServerMind helps

ServerMind installs on a homelab box in about two minutes and gives you a chat box plus a live dashboard. Ask about your stack in plain English; it runs the right read-only checks and explains what it finds.

Are all my containers up? How much disk is left on the NAS box? Is Pi-hole responding? Restart the media server What's hammering the CPU? Tail the docker logs for jellyfin

It manages the usual homelab cast — Docker, systemd services, PM2, Nginx, Caddy, Redis, MySQL/MariaDB and fail2ban — plus disk, memory, ports and logs. Add your own custom tools too: a frozen command for that one quirky service, an HTTP health check on a container, or a specific file to read.

Built for the self-hosting ethos

  • Fully offline-capable. Point it at a local Ollama model and nothing leaves your network — no cloud backend, no telemetry, zero data egress.
  • No domain, no exposure. It binds to localhost; reach it over an SSH tunnel or Tailscale. Got several machines? Agents dial OUT to the controller, so you open no inbound ports on any node — optionally over a self-hosted WireGuard mesh with no third-party coordinator.
  • Free and yours. MIT open-source — fork it, audit it, run it forever. No accounts, no licence checks.
  • Safe to leave on. Read-only allowlist by default; restarts and stops need the arm switch you control.

One Pi or a rack of them

A single machine runs zero-config. When your lab grows to several nodes, designate one as a controller and run lightweight agents on the rest — then ask one question across everything: "which boxes are low on disk?" Reports and alerts from every node land in one inbox, and you get a daily health digest plus alerts on high disk/memory, a downed service, or a TLS cert about to expire.

In one lineThe convenience of an AI ops assistant, with the privacy a homelab is supposed to give you.

FAQ

Can it run fully offline?

Yes — with a local Ollama model nothing leaves your network. There's no cloud backend and no telemetry.

Do I need a domain or to expose my lab?

No. It binds to localhost (reach it via SSH tunnel or Tailscale), and across multiple machines agents dial out, so no inbound ports are opened.

Does it handle Docker and friends?

Yes — Docker, systemd, PM2, Nginx, Caddy, Redis, MySQL/MariaDB, fail2ban, plus disk, memory, ports and logs, and your own custom tools.

Get started

$ curl -fsSL https://servermind.dev/install.sh | bash

Related use cases: Indie developers · On-call & incident response · Safe database operations

← All use cases  ·  Read the setup guide →