Documentation
One account, one API key, one credit balance — and 17 calibrated prediction APIs. Here's how to make your first call.
Authorization: Bearer <key>.Every request needs your API key in the Authorization header. No cookies, no sessions —
just a bearer token.
curl -X POST https://api.dryapi.dev/predict \
-H "Authorization: Bearer $DRYAPI_KEY" \
-H "Content-Type: application/json" \
-d '{{"model": "property/valuation/v1", "features": {{"bedrooms": 3, "bathrooms": 2, "sqft_living": 2000, "year_built": 1990, "zip_code": "98115"}}}}' Send a model id and a flat features map matching that model's schema.
{{
"model": "property/valuation/v1",
"features": {{
"bedrooms": 3,
"sqft_living": 2000,
"property_type": "SingleFamily",
...
}}
}} Every response includes calibrated confidence intervals, a confidence score, and top influencing factors.
{{
"value": 722396.99,
"interval": [670715.79, 774078.19],
"confidence_interval": {{"lower": 670715.79, "upper": 774078.19, "coverage": 0.9}},
"confidence": {{"level": 0.9, "score": 0.93}},
"top_factors": [{{"feature": "lot_acres", "impact": 12129}}, ...]
}} Opt in to training on your request by passing train_consent: true. Each consented call costs half the credits of a normal call, and your features help keep the model drift-free.
No personal data is stored.
{{
"model": "property/valuation/v1",
"features": {{ ... }},
"train_consent": true
}} Every prediction returns a prediction_id and a feedback_url. Send the actual
outcome back to POST /predictions/{id}/feedback — include the actual value or a
domain-specific field (e.g. sale_price for property/valuation/v1). Ground-truth records
earn 2 credits back per submission.
curl -X POST https://api.dryapi.dev/predictions/{prediction_id}/feedback \
-H "Authorization: Bearer $DRYAPI_KEY" \
-H "Content-Type: application/json" \
-d '{{"actual_value": 495000, "outcome_date": "2026-08-01"}}' Browse the full catalog or jump straight to a model. For the interactive reference, open the live API explorer at https://api.dryapi.dev/docs.