Skip to content
Laptop showing the Flash weather map with storm cells tracked over the southeast, the same model run the Flash API serves

Flash API · REST and webhooks · Same run as the app

The weather API that fires before the weather arrives.

One GET returns over 100 parameters for any 1×1 km cell — lightning risk, hail lead time, WBGT, gusts, rain rate and frost probability — with forecasts out to 180 hours, refreshed as often as every 2 minutes. Subscribe a webhook and your PA system, scheduler or dashboard hears about the storm before it lands.

For construction, schools, golf, utilities and the tools they already run · same model run as the Command Center and Skye · no hardware

Illustrative example · Not live weather

GEThttps://api.flashweather.ai/v1/cells/3391/now200 OK
{
  "cell_id": "3391",
  "run": "2026-09-21T12:02:00Z",
  "centroid": [34.0234, -84.3616],
  "lightning_risk_60m": 0.82,
  "lightning_first_strike_eta_min": 38,
  "hail_lead_min": null,
  "wbgt_f": 88.1,
  "wind_gust_mph": 31,
  "rain_rate_in_hr": 0.60,
  "frost_prob": 0.00,
  "next_alert": "lightning.warning @ T-15"
}
WEBHOOKPOST https://hooks.yourapp.com/flash · event: lightning.warning
{ "event": "lightning.warning", "site": "Practice Field B",
  "cell_id": "3391", "eta_min": 15, "risk": 0.82,
  "issued": "2026-09-21T14:21:00Z", "signature": "sha256=…" }

Endpoints · REST · JSON

Which endpoints does the API expose?

The resources cover cells, your sites, alert routing, regional outlooks and the validation dataset behind the accuracy figure. Authentication, pagination and versioning are in the developer docs.
Flash API endpoints, what each returns and typical uses
EndpointReturnsTypical use
Current risk and readings for one 1×1 km cell, with the model run it came fromPA triggers, site dashboards, kiosk displays
GET/v1/cells/3391/now200 OK · illustrative
{
  "cell_id": "3391",
  "run": "2026-09-21T12:02:00Z",
  "lightning_risk_60m": 0.82,
  "lightning_first_strike_eta_min": 38,
  "wbgt_f": 88.1,
  "wind_gust_mph": 31
}
Hourly forecast out to 180 hours for every parameter on the cellScheduling, crew plans, spray windows
Your registered sites, their cells, thresholds and channelsSync with your asset list or ERP
Subscribe a threshold per site to a channel or a webhook URLAlert routing, horn and strobe relays, Slack and Teams
Regional outlook for one layer: wbgt, wind, rain, frost, hail, lightningPlanning views across a region or a portfolio
The NLDN-scored validation dataset behind the up to 99.6% figure on the 60-minute windowYour own audit; see the accuracy method

Parameters · A sample of over 100

What can you request for a cell?

Every parameter is served per 1×1 km cell, per hour, from the same run. Units, ranges and the full list live in the developer docs so they never drift from the response.
GET/v1/cells/2214/now?params=lightning_risk_60m,wbgt_f

2 of 100+ parameters selected

A sample of Flash API parameters, grouped by hazard. Each parameter is a toggle that adds it to the request above.
GroupParameters
Severe
Heat
Water
Wind
Agronomy
Visibility and winter

Webhooks · Per site, per event, per URL

Which events can a webhook fire on?

Subscribe per site, per event, per URL. Every payload carries the cell, the model run and a signature, and every delivery is logged so your on-call engineer can replay it.
Flash API webhook events, when each fires and its lead time. Choose an event to log a sample delivery.
EventFires whenLead
A cell inside the site radius enters the 60-minute windowT–60
Strike-level prediction for the site's own cellConfigurable
Your all-clear timer completes with no new cell inside the radiusAfter the last strike
A Predictive Hail cell with a size class is heading for the siteUp to T–55
The forecast will cross 82, 87, 90 or 92 °F, or your own bandHours ahead
The first gust over the site's crane, lift or tent limitGust timing
Frost or freeze probability over your threshold for a fieldHours ahead

Delivery log

Illustrative example · not live weather

  1. 13:24:02lightning.watchPractice Field B · delivered ✓ signed
  2. 14:21:00lightning.warningPractice Field B · delivered ✓ signed
  3. 14:29:40hail.approachingPractice Field B · delivered ✓ signed
  4. 15:12:06lightning.all_clearPractice Field B · delivered ✓ signed

POST hooks.yourapp.com/flash

Rate limits · SLAs · Published once

What are the rate limits and SLAs?

We publish limits and commitments in one place, the developer docs and your order form, so a marketing page never quotes a number that has since changed.
  • Rate limits

    Per key and per minute, with burst allowances for alert fan-out. The current figures are published in the developer docs.

  • Uptime and support

    The uptime commitment, support hours and escalation path are written into your order form and mirrored in the developer docs.

  • Webhook delivery

    Signed payloads, retries with backoff and a replay endpoint. The retry schedule is published in the developer docs.

See current limits and SLAs in the developer docs (opens in a new tab)

An abstract network of glowing nodes joined by thin lines in dark space, standing for the Flash API delivering the same forecast to every connected system

One contract · Every cell

Over 100 parameters, forecasts out to 180 hours, refreshed as often as every two minutes.

Skye · Illustrative example · Not live weather

Build with Skye

Skye, Flash Weather's AI weather agent, reads the same endpoints you do, and your app can call the agent. Send a plain-language question with a site scope; get back an answer, the cells and run it read, and any action it proposes, held until a person confirms.

Permissions per connector · every action logged · human confirmation before any schedule or record changes.

Skye
A darkened operations control room at night with a wall of screens casting blue glow across empty desks, where Flash API responses and Skye's answers land side by side

One engine · Two front doors

Your app calls the API. Your crew asks the agent.
POSThttps://api.flashweather.ai/v1/agent/ask200 OK · illustrative
curl -X POST https://api.flashweather.ai/v1/agent/ask \
  -H "Authorization: Bearer $FLASH_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "question": "Which of my sites are inside a lightning window before 3 pm?",
        "scope": "sites:all", "actions": "propose" }'
const res = await fetch("https://api.flashweather.ai/v1/agent/ask", {
  method: "POST",
  headers: {
    Authorization: `Bearer ${process.env.FLASH_API_KEY}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    question: "Which of my sites are inside a lightning window before 3 pm?",
    scope: "sites:all",
    actions: "propose",
  }),
});
const answer = await res.json();
import os
import requests

res = requests.post(
    "https://api.flashweather.ai/v1/agent/ask",
    headers={"Authorization": f"Bearer {os.environ['FLASH_API_KEY']}"},
    json={
        "question": "Which of my sites are inside a lightning window before 3 pm?",
        "scope": "sites:all",
        "actions": "propose",
    },
)
answer = res.json()

Response

{ "answer": "Two: Midtown Parking Deck (cell 3391, T–38) and Riverside Yard (cell 3402, T–52).",  "cells": ["3391", "3402"], "run": "2026-09-21T12:02:00Z",  "proposed_action": { "tool": "procore", "type": "reschedule_lift",    "status": "awaiting_confirmation" } }

A person confirms before anything changes.

Integrations · The tools you already run

Where does Flash already plug in?

Native connectors for the calendar, CRM, ERP and job-site tools our customers run, plus a signed webhook for everything else. Each connector is permissioned on its own.
  • Google Calendar

    Practice and shift moves

    Receives:lightning.watchwbgt.threshold

  • Microsoft 365

    Outlook, SharePoint

    Receives:lightning.warninglightning.all_clear

  • Salesforce

    Site records, field service orders

    Receives:hail.approachingwind.gust_over_limit

  • HubSpot

    Canvass lists, claims follow-up

    Receives:hail.approaching

  • NetSuite

    Work orders, fleet tasks

    Receives:wind.gust_over_limitfrost.expected

  • SAP

    Plant maintenance notifications

    Receives:lightning.warningwind.gust_over_limit

  • Procore

    Lift schedules, daily logs

    Receives:wind.gust_over_limitlightning.warning

  • Slack

    A channel per site, alerts and all-clears

    Receives:lightning.warninglightning.all_clear

  • Microsoft Teams

    Ask Flash from the phone

    Receives:lightning.watchwbgt.threshold

  • Any webhook

    PA systems, horn relays, your own stack

    Receives:every event

Agentic

Skye

Ask in plain language; the agent calls the same API and acts in your tools. Flash provides the harness: the prediction engine, your sites and your data, connected to the tools you already run. Ask in plain language; get an answer or an action.

Skye

The same engine · Other ways to take delivery

  • Flash Edge Model product render: three national ensemble maps stacked above the Flash Edge wordmark, shown as another way to take delivery of the same engine

    Ensemble

    Flash Edge Model

    Best, most likely and worst case for over 100 parameters, so you plan for the range, not a single line.

    See Flash Edge Model
  • The Weather Command Center map: lightning probability across north and east Texas at 1×1 km, a band of storm cells from Fort Worth to Texarkana shaded yellow to red, with site temperature markers

    Platform

    Weather Command Center

    Flash's flagship platform: every risk that touches your operation on one screen.

    See Weather Command Center
  • Flash Mobile App product render: four phone screens showing lightning probability, future radar, live radar and seven-day forecasts

    Mobile

    Flash Mobile App

    Enterprise-grade prediction in your pocket, with 18 hours of future radar.

    See Flash Mobile App

FAQ

What do developers ask before they integrate?

The four questions from every integration call, answered with the same numbers as the rest of the site.

Yes. The API, the Command Center, the mobile app and Skye read the same 1×1 km cells from the same run, so a webhook and a push alert never disagree about a site.

The lightning model refreshes every 2 minutes and the hail model every 5. Heat, wind, rain and the agronomy parameters are forecast hourly, out to 180 hours for planning. Every response carries its run timestamp.

Yes. /v1/validation/lightning returns the NLDN-scored dataset behind the up to 99.6% figure on the one-hour window, so you can re-run the scoring yourself. The method is written up on the accuracy page.

Yes, through the agent endpoint. It answers with the cells and the run it read; any action it proposes waits for a person to confirm, and each connector is permissioned on its own, so your app decides what the agent may touch.

Book a demo · No sensors required for the forecast

See your own sites on the map before the next storm.

We'll load your locations, replay last season's strikes against Flash's predictions, and show you the alert log your safety officer would have received.