Skip to main content
How AI agents manage homelab services, and what’s not yet automatable.

Approach

Each service directory has an AGENTS.md describing its management interface. Agents discover ports, URLs, and env var names at runtime from compose files and .env files — nothing is hardcoded. Preference order:
  1. CLI via docker exec or host-installed tools
  2. REST API via curl
  3. Config file edits + container restart

Host Dev Environment

Node.js is managed via fnm (installed to ~/.local/bin), keeping the OS clean. Node versions and global packages live in ~/.local/share/fnm/. Globally installed tools:
  • wrangler — Cloudflare CLI (Workers, DNS, tunnels, etc.)
  • kuma — Uptime Kuma CLI (monitors, notifications, tags, maintenance)

Gaps

Services that can’t be fully managed by an agent today:
ServiceIssueWorkaround
MinifluxNo API key configuredGenerate one in UI: Settings → API Keys
Uptime KumaSession-cookie auth only, no API keykuma CLI installed on host
YamtrackSession-cookie auth only, no API keyNone — UI only
ChaptarrNo API or CLIUI only
Draw (Excalidraw)No API or CLIUI only, data in filesystem
TwakeNo API or CLIUI only