LB-06 Incident Commander

A team of AI agents diagnoses a live incident in a simulated microservice stack and proposes a fix for you to approve.

LB-06

Incident Commander Evaluation board

Live

A team of AI agents diagnoses an incident in a simulated shop. Break the shop with one of four faults, watch the dashboards turn, follow what the agents find, and approve the fix they propose. Nothing is changed without your click, and the incident closes only when the shop has really recovered.

The shop is a simulation. One minute of its time passes every 2 seconds, an incident ends on its own after 8 minutes of real time, and the same seed always makes the same incident.

Break the shop

Choose what to breakThe cart release that broke checkoutLook for a deploy that came right before the errors.Bad deployThe slow payment providerLook for latency that starts in one service and spreads.Slow payment providerThe inventory memory leakLook for memory that climbs minute after minute.Memory leakThe cache stampedeLook for load on the database that follows the cache.Cache stampede

Running it live uses your one incident today. The agents make nine to fifteen model calls, and none when the same incident has been run before.

Make your own incident

Limits

Incidents left todayCounting…

Operating limit	Value
Incidents per visitor per day	1
Step cap per incident	15
Remediation	Human approval
Model calls per incident (est.)	9–15

Your incident today

You have not started an incident today.

The Scope

Start a run and its trace appears here, step by step.