Loading…
FortyGuard API → Fetch Temperature → Sentinel Anomaly Detection → Heat-Risk Scoring → Alerts + Recommendations → MCP Tool Interface

Monitoring Stats

—
Readings
—
Active Alerts
—
Scans Run
—
Monitored Cities

City Temperature Grid

Click "Scan All Cities" to fetch live temperatures.

Active Heat-Risk Alerts

No alerts loaded. Click Refresh.

Single-City Deep Monitor

MCP Tool Interface

Thermal Sentinel exposes 11 MCP tools. Any MCP-compatible AI agent can call these to monitor heat risk, detect anomalies, and generate urban-planning recommendations.

Try a Tool

// Select a tool and execute.

About Thermal Sentinel

Thermal Sentinel is an AI agent built for the FortyGuard Hackathon'26 (Agentic AI track). It uses the FortyGuard Temperature API to monitor real-time urban temperatures, detects heat waves and temperature anomalies using Sentinel-style anomaly detection, and generates heat-risk alerts with urban-planning recommendations.

Judging criteria alignment:
• Impact (40%) — Heat waves kill more people than any other weather event. Real-time detection + actionable urban-planning recs save lives.
• Technical Execution (35%) — FastAPI backend, MCP protocol, multi-signal anomaly detection (WHO bands + z-score + sustained-heat + rate-of-change), composite 0-100 risk score, file-based persistence for serverless.
• Innovation (15%) — Sentinel-pattern anomaly detection applied to climate/heat data; MCP tool surface lets any AI agent monitor heat risk.
• Communication (10%) — Live dashboard + one-click demo + narrated summary.

Architecture

┌─────────────────────────────────────────────────────────┐
│                    Dashboard UI (static)                 │
│              REST API  +  MCP Tool Interface             │
├─────────────────────────────────────────────────────────┤
│                     FastAPI (app/main.py)                │
│         /api/* routes  +  /mcp/tools  /mcp/call          │
├─────────────────────────────────────────────────────────┤
│                    Agent (app/agent.py)                  │
│     scan_cities → monitor_city → heat_risk_report        │
├──────────────┬───────────────────┬──────────────────────┤
│ FortyGuard   │  Anomaly Detector │     Store (SQLite)    │
│  Client      │  (Sentinel-style) │   /tmp persistence    │
│              │                   │                      │
│ Real API or  │  WHO heat bands    │  readings + alerts   │
│ mock data    │  Z-score outlier   │  + scans tables      │
│              │  Heat-wave detect  │                      │
│              │  Rate-of-change    │                      │
│              │  Composite score   │                      │
└──────────────┴───────────────────┴──────────────────────┘

What's Real vs Mocked

  • Real: FortyGuard API integration (when FORTYGUARD_API_KEY set), anomaly detection engine, MCP tool protocol, alert generation, dashboard UI, Vercel deployment.
  • Mock (fallback): When no API key, temperature data is generated from per-city climate baselines with realistic diurnal variation and anomaly injection.