Home & self-hosted

Run your self-hosted home lab by chat

Q: How do I manage Docker containers in my home lab remotely?

Install the AI Commander agent on the host running Docker, then ask your AI client to run docker commands by quoting the machine's session code — for example pull a new image and restart a container. Output streams back as text, so the agent can confirm the container is healthy.

A home lab is a pile of services you keep healthy: Docker stacks, a media server, Pi-hole, *arr apps, a backup job. AI Commander lets you operate all of it by asking Claude in plain language — pull an image, restart a container, check why something's unhealthy — without SSHing in or exposing anything.

The job

Home-labbers love tools like Tailscale because they make a private box reachable without opening ports. AI Commander complements that: instead of reaching the box and then typing commands yourself, you hand the shell to an AI client that runs the commands for you and reads the output back. Typical chores:

Update a container: docker compose pull && docker compose up -d, then confirm it's healthy.
Investigate an unhealthy service: read docker logs, spot the error, fix the env var.
Reclaim space: docker image prune, rotate logs, check volume usage.
Update Pi-hole's blocklists or restart pihole-FTL when DNS gets flaky.

What it looks like

Claude — MCP

You: on home-lab, update the jellyfin stack and tell me if it came back healthy Claude: > cd /srv/jellyfin && docker compose pull jellyfin Pulled (10.9.11) > docker compose up -d > docker inspect -f '{{.State.Health.Status}}' jellyfin healthy Claude: Updated to 10.9.11 and the container reports healthy.

AI Commander vs Tailscale for a home lab

They overlap but aren't the same tool. A quick orientation:

	AI Commander	Tailscale + SSH
What it gives you	A shell an AI client can run	A private mesh network
You type the commands	AI does, you describe	Yes, you do
Inbound ports	None	None
MCP / AI native	✓	✗
Good for	Operating a box by chat	Reaching everything on your tailnet

Use both. Tailscale for general network reach, AI Commander when you'd rather say "fix it" than SSH in and type.

Set it up

On your home-lab host (Linux), install the agent:

curl -fsSL https://aicommander.dev/install | sudo bash

It runs as a systemd service and prints a stable session code. Connect your AI client, quote the code, and start asking. Running several boxes? Sign in to give each a friendly alias.

FAQ

How do I manage Docker containers in my home lab remotely?

Install the agent on the Docker host, then ask your AI client to run docker commands by quoting the session code — e.g. pull a new image and restart a container. Output streams back as text so the agent can confirm the container is healthy.

Is AI Commander an alternative to Tailscale for a home lab?

It solves a related problem differently. Tailscale builds a mesh VPN so your devices reach each other; AI Commander gives an AI client a shell on a specific machine via an outbound-only agent. Many home-labbers use both.

Do I need to expose my home lab to the internet?

No. The agent connects outbound to the relay, so no inbound ports are opened and nothing is exposed publicly.

Hand your home lab to an AI shell

Install the agent on one host and ask Claude to tidy up your containers tonight.

Install the agent Connect your AI client