Intake
Evidence in. Root cause out.
Diagnose what broke — and watch what might break next.
MADD correlates tickets, logs, traces, Maximo configuration, runtime signals, recent changes and version-breaking changes. The pilot never executes corrective actions.
Observer online7 streams · 34 probes · upgrade watch active
Local simulation
Investigation
Waiting for incident
Select a scenario and run the investigation.
Exception rate2/daybaseline
Integration backlog142stable
Manage restarts024h
DB2 lock waits3baseline
JVM GC p9542msbaseline
Anomaly score0.08normal
Live simulator
Environment stream
Pre-emptive diagnosis
Signals worth investigating
No abnormal signals yet.
Past + future breaking change watch
Version exposure
Continuous intelligence
What MADD watches
Explainable diagnosis
Evidence graph
Ticket-created webhook → classify → investigate → append evidence and status.
incident.read · comment.writeGET APIs, metadata queries, approved SQL SELECT probes, logs and system information.
read-only capability contractPod status, events, restarts, logs, metrics and OpenTelemetry streams.
get/list/watch onlyPilot boundary
No generic shell. No arbitrary SQL. No remediation tool.
The agent chooses from explicit read-only probes. Real customer connectors replace the simulator adapters without changing the investigation model.