Roja is a lightweight agent that runs inside your network, watches your devices, and closes the loop between an alert and the fix. Here's the whole loop, honestly described.
The agent polls and listens to your devices directly — interface state, routing adjacencies, control-plane health, resource pressure. It detects incidents from device state, not from a downstream ticket. When something degrades, Roja knows in seconds.
Roja works the problem the way an engineer would: it reads the relevant state from the affected devices, correlates symptoms across neighbors, and isolates a root cause. A flapping BGP session, a loop introduced by a rogue switch, an exhausted DHCP scope — each produces a diagnosis you can read in plain language, with the evidence attached.
This is where Roja is different, and where we're most careful.
Every action Roja can take is pre-authorized, cryptographically verified, and validated before it executes. The agent does not run arbitrary commands — not ones it invents, not ones anyone else injects. If the fix is safe and pre-authorized, Roja applies it autonomously. If a change carries risk, Roja stops and asks a human first, with the diagnosis and proposed fix laid out for one-click approval.
Before any change, Roja captures device state. If a fix doesn't take, it rolls back to exactly where you were.
A fix isn't done when the command runs — it's done when the network recovers. Roja re-checks the original symptoms, confirms steady state, and only then closes the incident. Every step lands in an append-only audit log: what was detected, what was diagnosed, what ran, what changed, who (or what) approved it.
The agent initiates all connectivity outbound. Your device credentials stay in your environment — we never host them. The safety framework sits between every decision and its execution.
Patent pending.
