CrisisMind AI

Most AI tools only answer questions. In a real crisis, people don't need Q&A — they need a plan. CrisisMind AI takes a single sentence like "There is a flood in Mumbai, what should I do?" and turns it into a complete, actionable response. Behind the scenes, seven specialist AI agents run in sequence, each handing off structured work to the next: a Planner classifies the crisis type and location; a Research agent pulls live advisories and weather data; a Risk agent scores severity (LOW to CRITICAL); a Response agent builds a prioritized action plan and emergency checklist; a Communication agent drafts a ready-to-send SMS and email so users can tell family or employers they're safe without composing anything under stress; a Summary agent compresses everything into a 10-second situation brief; and a Validator agent checks the output for gaps before it reaches the user. The pipeline isn't hidden behind a spinner — the frontend shows each agent lighting up live as it works, making the "agentic" behavior visible rather than a black box. CrisisMind AI also adapts its playbook by crisis type: natural disasters (floods, earthquakes, wildfires, cyclones) get evacuation-and-safety guidance, while a ransomware report instead produces a security incident response plan — disconnect infected devices, don't pay the ransom, preserve logs, notify IT and CERT. Built with FastAPI and LangGraph on the backend for multi-agent orchestration, Fireworks AI for reasoning, Tavily for live search grounding, Supabase for persistence, and a React + Tailwind frontend with a mission-control-style UI. Every report can be exported as a PDF and revisited later in a searchable history and dashboard.