us-property-hazard-risk (MCP server)

A remote MCP server that assesses natural hazard exposure — flood, wildfire, earthquake, and coastal proximity — for a single US property address, using only free, keyless public data. No API keys, no database, no map. Works from Claude, ChatGPT, or any other MCP-compatible host.

Tool

assess_property_hazard(address: str)

Validates and geocodes the address, then returns flood zone (FEMA NFHL), wildfire hazard score, seismic PGA (USGS, ASCE 7-22), fault and coastal distance, and a deterministic 0–100 hazard_score / hazard_tier with up to 5 hazard_drivers. If a data source is unavailable, that field is reported as unknown rather than assumed zero-risk.

Data sources

• US Census Bureau Geocoder — address validation + geocoding
• FEMA NFHL — flood zone
• USGS Design Maps — seismic PGA
• Static reference points (haversine) — fault + coastal proximity

Local dev

python -m venv venv
venv/Scripts/pip install -r requirements.txt   # venv/bin/pip on macOS/Linux

cd src
../venv/Scripts/python server.py               # ../venv/bin/python on macOS/Linux

Starts a streamable-HTTP MCP server on http://localhost:8000/mcp:

curl -X POST http://localhost:8000/mcp \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"assess_property_hazard","arguments":{"address":"1600 Pennsylvania Ave NW, Washington, DC 20500"}}}'

Deployment

Runs via Docker Compose (docker-compose.yml):

docker compose up -d --build

Known limitations

• US addresses only
• FEMA NFHL is occasionally unreachable from some networks (flood_available: false)
• Wildfire scoring is a coarse coordinate-zone heuristic, not parcel-level

License

MIT — see LICENSE.