Homebutler is a new open-source tool that flips the script on AI-driven server management. Instead of handing your favorite LLM an SSH key and hoping for the best, this single Go binary exposes narrow, read-heavy operations through MCP (Model Context Protocol)—the same protocol powering Cursor, Claude Desktop, and a growing ecosystem of AI-native tooling.

How It Works

The tool maps your homelab topology first: containers, exposed ports, system listeners, and their relationships. Once Homebutler knows what exists, it serves that context to your AI assistant through MCP tools like inventory reporting, Docker status checks, backup verification drills, and disk space monitoring. When something goes wrong at 3 AM—a crashed container, a filling disk—your AI agent can diagnose and act using scoped operations rather than dropping you into a full shell.

Narrow Tools Over Full Access

The pitch is explicit: your AI gets narrow tools instead of a full bash prompt. Homebutler surfaces crash evidence before logs rotate out, generates Mermaid diagrams for GitHub issues or Obsidian notes, and connects Docker-published ports back to the containers that own them. Public versus local port exposure becomes immediately visible—no more guessing which service is listening where.

Security Model

By avoiding SSH entirely, Homebutler sidesteps a whole class of credential management problems. There's no private key to rotate, no ~/.ssh/authorized_keys to maintain across machines. The binary runs locally and your AI talks to it over MCP—a pattern that keeps blast radius contained even if an agent goes sideways on you.

One-Command App Deployment

Beyond monitoring, Homebutler ships with one-command installs for 15+ self-hosted apps: Uptime Kuma, Vaultwarden, Pi-hole, Jellyfin, Plex, Portainer, Nginx Proxy Manager, Nextcloud, Home Assistant, Grafana, Prometheus, WireGuard, Immich, FreshRSS, and more. Spin up a fresh homelab app in seconds without hunting down docker-compose files.

Coming Soon: Autonomous Nightly Maintenance

The roadmap includes AI agents that run health checks every night automatically—pruning disks when they fill, restarting crashed containers, refreshing expired backups—all while you sleep. You wake up to a single summary line: "All systems nominal." The rule shaper config shown in the docs suggests human approval gates for destructive actions like disk cleanup, keeping automation explainable before it becomes unconstrained.

Key Takeaways

  • Single Go binary with no external dependencies beyond Docker
  • MCP integration means your AI gets structured context before acting—no blind commands
  • Backup Drill feature boots restores in isolated containers to prove they actually work
  • Web dashboard included for human-in-the-loop monitoring
  • SSH never touches the equation, reducing credential sprawl

The Bottom Line

Homebutler is exactly the kind of pragmatic security thinking the homelab world needs right now. As AI agents get more capable and more people experiment with autonomous ops, tools that constrain blast radius without sacrificing functionality will become essential infrastructure—not optional nice-to-haves.