The strongest part is connecting diagnosis with action instead of making the monitoring dashboard another place to look before opening SSH. The confirm-gated controls also seem like a sensible balance between convenience and giving a monitoring tool too much authority.
Really appreciate that — the “diagnosis → action” bit was actually the hardest part to get right in design. It’s easy to build a dashboard that just shows you more graphs, but the whole point was to collapse the “notice something’s wrong → SSH in → poke around → run the fix” loop into one place. If it still ends with you tabbing over to a terminal, it’s not really saving you anything.
The confirm-gates came out of being paranoid about the opposite failure mode — a monitoring tool that can silently restart services or kill processes on its own is a liability waiting to happen. So the rule I landed on: it can suggest and stage the fix instantly, but a human always pulls the trigger. Glad that balance reads as sensible from the outside too.
That’s an interesting balance. The fact that the tool can prepare the fix without taking the final action seems particularly important for something with that level of access.
About
VPS monitoring that lets you actually fix things, not just watch them break. Every tool I tried stopped at a graph and made me SSH in anyway. This one restarts services, kills processes, and reboots from the dashboard.
3 Comments
The strongest part is connecting diagnosis with action instead of making the monitoring dashboard another place to look before opening SSH. The confirm-gated controls also seem like a sensible balance between convenience and giving a monitoring tool too much authority.
Really appreciate that — the “diagnosis → action” bit was actually the hardest part to get right in design. It’s easy to build a dashboard that just shows you more graphs, but the whole point was to collapse the “notice something’s wrong → SSH in → poke around → run the fix” loop into one place. If it still ends with you tabbing over to a terminal, it’s not really saving you anything.
The confirm-gates came out of being paranoid about the opposite failure mode — a monitoring tool that can silently restart services or kill processes on its own is a liability waiting to happen. So the rule I landed on: it can suggest and stage the fix instantly, but a human always pulls the trigger. Glad that balance reads as sensible from the outside too.
That’s an interesting balance. The fact that the tool can prepare the fix without taking the final action seems particularly important for something with that level of access.