Pay for the quiet, not the volume.
Start free. Upgrade when your stack outgrows the limit. No seats, no sales calls.
Free
One homelab, watched quietly.
- 500 triaged events / month
- 5 GB searchable, 24-hour retention
- All channels, all suppression
- 10 assistant chats / month
- Unlimited MCP usage
Pro
Every host, two weeks of history, room for noise.
- 10,000 triaged events / month
- Full investigations — dig deeper, get a fix
- 25 GB searchable, 14-day retention
- Multiple Dozzle instances
- 200 assistant chats / month
- Unlimited MCP usage
- Priority support
Team
Headroom for a real production stack.
- 50,000 triaged events / month
- Full investigations — dig deeper, get a fix
- 100 GB searchable, 30-day retention
- 1,000 assistant chats / month
- Unlimited MCP usage
What changes with each tier.
| Limit | Free | Pro | Team |
|---|---|---|---|
| Triaged events / month | 500 | 10K | 50K |
| Standard triage (what broke, kept quiet) | Yes | Yes | Yes |
| Full investigation (deeper dive + a fix) | — | Included | Included |
| Searchable log volume | 5 GB | 25 GB | 100 GB |
| Retention | 24 hours | 14 days | 30 days |
| Connected Dozzle instances | 1 | Unlimited | Unlimited |
| Assistant chats / month | 10 | 200 | 1,000 |
| MCP server access | Unlimited | Unlimited | Unlimited |
| Priority support | — | Yes | Yes |
Suppression, summaries, search, every channel, and remote control are on every plan.
What's a Full investigation, and how is it different from triage?
Triage runs on every plan: it watches your containers, groups the noise, tells you what broke, and keeps you quiet when nothing's wrong. It takes a fast, bounded look — and flags when it couldn't pin the cause down. A Full investigation (Pro and Team) is the deep dive for exactly those cases: one click and it keeps digging — more passes, a stronger model, correlating across your containers, hosts and timeline — then hands you concrete steps to fix it. Same idea, taken further, with a fix at the end.
What counts as a triaged event?
Any container event or matching log line that's fed through the triage pipeline: where we figure out whether to send a new alert, fold it into an existing one, or stay silent. You're paying for that work, not for raw storage.
What happens if I go over?
Nothing breaks. You drop into sampling mode: triage pauses, your event history keeps recording, and roughly one in ten events comes through as a raw alert so you can still see what's happening. You'll feel it in your inbox before you see it on a usage page.
If a container exits 47 times, is that 47 events?
Yes against the limit. But you get one alert that says 47, not 47 alerts. That's the whole point.
Will 500 triaged events be enough?
Most single-homelab setups (Sonarr, Radarr, Immich, NPM, the usual stack) sit well under it. Watchtower or a heavy logger will push you over. Upgrade anytime, no data lost.