Send your sensor readings, metrics, or KPIs — get back anomalies, expected values, and severity. No data science team, no ML pipeline, no US cloud required.
# Same request body as Azure Anomaly Detector curl -X POST https://api.anomatik.de/anomalydetector/v1.1/timeseries/entire/detect \ -H "Ocp-Apim-Subscription-Key: $ANOMATIK_KEY" \ -H "Content-Type: application/json" \ -d '{ "granularity": "hourly", "series": [ { "timestamp": "2026-07-14T00:00:00Z", "value": 21.5 }, { "timestamp": "2026-07-14T01:00:00Z", "value": 22.1 }, { "timestamp": "2026-07-14T02:00:00Z", "value": 97.4 } ] }' # → { "isAnomaly": [false, false, true], "severity": [0, 0, 0.94], ... }
Anomatik was built by an IoT & network engineer — for sensor fleets, lab equipment, energy meters, and application metrics.
Same endpoints, same JSON schema, same auth header as Azure Anomaly Detector v1.1. Migration = swap the URL and the API key. Nothing else changes.
Hosted exclusively in EU data centers. GDPR-friendly by architecture: your data is processed in memory and never stored. DPA available on request.
Robust to noise, gaps, and irregular readings — vibration, power, temperature, occupancy. Handles seasonality (daily/weekly patterns) automatically.
No model training, no configuration. Send at least 12 data points, get results back in milliseconds. Copy-paste examples for Python, JavaScript, and curl.
Every anomaly comes with the expected value, upper/lower margins, and a severity score — so you can tune alerting instead of guessing.
A generous free tier and flat monthly plans. No capacity units, no per-node pricing, no surprise bills at the end of the month.
Microsoft retires the service on October 1, 2026. Here is your entire migration:
Replace *.cognitiveservices.azure.com with api.anomatik.de. The paths /anomalydetector/v1.1/timeseries/entire/detect and /last/detect stay identical.
Your existing Ocp-Apim-Subscription-Key header keeps working — just use your Anomatik key.
Request and response schemas are compatible. Your parsing code, alerts, and dashboards keep working unchanged.
Simple plans from day one. Early-access members get 50% off for the first 12 months.
The private beta starts before the Azure retirement date. Waitlist members get priority onboarding, free migration help, and 50% off for 12 months.
No spam. One email when the beta opens, one before Azure retires. Unsubscribe anytime.
Yes — the univariate v1.1 endpoints (/timeseries/entire/detect and /timeseries/last/detect) accept the same request body and return the same response fields (expectedValues, upperMargins, lowerMargins, isAnomaly, severity, …). Multivariate detection is on the roadmap.
All processing happens in EU data centers. Time-series payloads are processed in memory and are not persisted. We store only your account data and usage counters for billing.
Any numeric time series with at least 12 points: sensor readings (temperature, power, vibration), infrastructure metrics, business KPIs. Seasonality (daily/weekly cycles) is detected automatically.
Private beta opens well before the Azure retirement date (October 1, 2026), so you have time to migrate and test. Waitlist members get access first.