Break something.
Watch the AI fix it.
Pick a failure and it gets injected into a simulated company network. The AI on-call engineer only sees the user's complaint. It has to reproduce the problem, follow the evidence to the cause, ask you before fixing anything and write up what happened.
- Not checked yet
- Healthy
- Degraded
- Failing
Choose what to break
The servers, logs and changes here are all simulated. Every fix waits for your approval, and the agent only sees the network through the same tools an on-call engineer would use.