# Atlas United Address-Level Storm Data API

Atlas United builds risk models for every U.S. address. This API sells five products, pay per address, no contract, with a free preview on every dataset:

1. ML Storm Forecast by Address (dataset=risk). Address-level forecasts for the next 24 hours to 10 days from a separate machine-learning model for each of 10 perils (tornado, hail, wind, flash flood, heavy rain, winter, ice, coastal surge, hurricane, wildfire). Windows 24h, 48h, 3d, 5d, 7d, 10d, refreshed 4 times a day. $0.05/address.
2. Model-Based Weather by Address (dataset=weather). Conditions, active severe weather alerts and a short-term rain outlook for a U.S. coordinate. Lean response free; details=true $0.0002/call.
3. Model-Based Hourly Forecast by Address (dataset=weather with hours_ahead=0-48). $0.0002/call with details=true.
4. Storm History by Address: Reported and Radar-Detected Events Since 2003 (dataset=history, summary in dataset=hazard_score). 11 event categories, one citable report. $0.05/line, $10 minimum; hazard_score $0.50/property.
5. Past Storm Leads by Address: Reported and Radar-Detected, Last 365 Days (dataset=storms). $0.05/address.

Forecasts describe what could happen, not what will happen. Storm history and past storm leads are records of reported and radar-detected events, not damage assessments.

## For AI agents

Call GET /v1/data with no parameters first. It returns the full, self-describing, machine-readable menu (every dataset, every field, every price, real examples). Trust that live response over this file if they ever disagree; this file is a pointer, not the source of truth.

This API also supports the x402 protocol for autonomous, per-request payment in USDC on Base network, no account, no signup, no human required. Machine-readable payment manifest:
https://data.truefixr.com/.well-known/x402

Human/API-key access (prepaid balance): https://truefixr.com/api or https://atlasunited.io/api
MCP server (Claude, ChatGPT, other MCP clients): https://mcp.atlasunited.io/mcp

## Keeping addresses out of URLs and logs

Send address and date of loss in a POST body to keep them out of URLs and logs. POST /v1/data takes a JSON object with the same parameters as the GET query string, with the same authentication, prices and free preview. GET with a query string keeps working. Example:

curl -X POST https://data.truefixr.com/v1/data -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_KEY" \
  -d '{"dataset":"history","address":"891 Lake Hollow Blvd SW, Marietta, GA 30064","date_of_loss":"2026-04-14","addresses":true}'

## Datasets (call the menu above for full detail on each)

- risk, ML Storm Forecast by Address: addresses in the forecast path of storms that could happen over the next 24 hours to 10 days. One record per address: full_address, city, state, zip, county, lat, lon, storm_type, forecast_time (UTC), risk_label, structure_value (USD, replacement cost, not market value), people (estimated occupants), facility_name, facility_type, storm_hits. $0.05/address. Free preview: counts and estimated cost (plus the county's A-F grade as a secondary label). Could happen, not will.
- weather, Model-Based Weather by Address and Model-Based Hourly Forecast by Address: free lean response; $0.0002/call for the full pull with details=true. Add hours_ahead=0-48 for forecast hours, or daily=true for a free 10-day daily outlook (date, high, low, condition, precipitation, wind).
- history, Storm History by Address: Reported and Radar-Detected Events Since 2003: 2003 to present, 11 event categories (hail, wind, flood, lightning, tornado, hurricane, winter storm, extreme heat/cold, drought, wildfire, coastal) in one report, with an optional date_of_loss= mode that lists reported and radar-detected events around that date (it does not confirm a claim or damage). $0.05/line, $10 minimum.
- hazard_score, Storm History Summary by Address: per-property summary by event category (frequency, most severe event, years since last event). $0.50/property.
- storms, Past Storm Leads by Address: Reported and Radar-Detected, Last 365 Days: one record per address with storm_type, event_time, severity and severity_unit, distance_mi and storm_hits. $0.05/address. A record of a reported or radar-detected event, not a damage assessment.
- address, Property Snapshot by Address: Past Storm Leads plus County Forecast Label: one property's own 365-day events (address-level match, about 150 m), with the county forecast label as secondary context. $0.05/property.
- facilities, ML Storm Forecast: Facilities at Risk, and at_risk, ML Storm Forecast: Schools and Businesses at Risk: facilities, schools and businesses inside the active forecast window. $0.05/record.
- wildfire, Wildfire Risk by County (county-level): county-level, one record per county, not address-level. $0.05/county.
- daily, Daily Forecast Digest by County (free), and coverage, Coverage Lookup (free): always free, no paid tier.

## Batch endpoints

- POST /v1/weather/batch, Model-Based Weather for Locations (batch): current conditions and active severe weather alerts for up to 500 locations in one call. $0.0002/location.
- POST /v1/portfolio/risk, County Forecast Status for Locations (batch): send up to 500 locations; per location returns county, window, the county forecast grade and a timestamp (county-level status, not address-level records). $0.05/location. For address-level forecast records use ML Storm Forecast by Address (dataset=risk).

## What this data is

Forecast records describe what could happen, not what will happen, and are not a damage prediction. Storm history and past storm leads are records of events reported or radar-detected at or near an address, not a property inspection, damage assessment or repair estimate, and they do not confirm damage or a claim. The county A-F grade is a whole-county label, secondary to the address-level records. Not affiliated with NOAA or the NWS.