Multi-agent system for crisis communication

Why they built it

To transition fragmented, slow manual disaster responses into coordinated, intelligent automated relief efforts using a scalable multi-agent architecture replacing complex microservices.

What worked

Parallel agent execution (<30s coordination), conditional severity-based logic, simplified architecture over microservices, sub-1s per agent/tool performance.

What broke or was painful

Not explicitly documented; relies on mock mode (USE_MOCK=1) for safe testing without live alerts

The result

<2 min end-to-end disaster response planning, allocated 12 resources, sent 1247 alerts across 3 demo scenarios