One call ahead.
POST your history. The forecast returns in 287 ms — confidence intervals included, best model auto-selected.
567,676,646 FORECASTS SERVED
History in.
Future out.
No training runs. No tuning. No ML team.
POST any series as JSON — sales, inventory, traffic, demand. No schema, no setup, no training.
Thirty-plus models compete on your data — ARIMA to Prophet to ETS. The most accurate wins, every call.
The forecast comes back with confidence intervals and the winning model's name. Median 287 ms.
curl -X POST https://forecastapi.com/v2/forecast \ -H "Authorization: Bearer $API_KEY" \ -H "Content-Type: application/json" \ -d '{ "data": [ {"date": "2024-01-01", "value": 152}, {"date": "2024-02-01", "value": 149}, {"date": "2024-03-01", "value": 164} ], "periods": 6, "frequency": "M" }'
{
"forecasts": [
{ "period": 1, "date": "2024-04-01",
"forecast": 177.1, "lower": 162.9, "upper": 191.4 },
{ "period": 2, "date": "2024-05-01",
"forecast": 179.4, "lower": 163.2, "upper": 195.6 }
],
"model_info": {
"method": "AutoETS",
"smape": 4.2,
"confidence": 0.95
}
}
No training.
No waiting.
There is no model to build. The first call is the deployment.
NO TRAINING PHASE · NO ENDPOINTS · NO MLOPS
Every model competes.
Each call runs your series against thirty-plus methods and scores them out-of-sample. You get the winner — and the scoreboard that proves it.
Uncertainty, quantified.
Every forecast ships with calibrated intervals, not a single guess. Plan on the median. Set safety stock on the upper bound. Sleep on the lower.
Sub-second, end to end.
The race runs in parallel on warm infrastructure — fast enough to sit inside a checkout, a dashboard, a replenishment loop. The timing breakdown ships in every response.
Different shapes.
Different models.
The API reads the shape of your series and picks accordingly.
Seasonal peaks, promotions, slow SKUs. Stock to the upper bound, not the guess.
MRR compounds; churn dampens. Project both, with intervals your CFO can plan on.
Hourly spikes on a weekly rhythm. Scale the fleet before the surge, not during it.
Intermittent, lumpy, zero-heavy series. Croston was built for exactly this.
"We deleted our forecasting pipeline and kept one API call."
One price.
Every model.
No tiers. No per-datapoint billing.
Volume pricing kicks in as you grow.
Stop guessing.
Your first 1,000 forecasts are free.
The first one is 287 ms away.