Clinical reasoning, in your product.
Roshan AI APIs return predictions, the concepts that produced them, and the verbatim phrases they're grounded in. One auth flow, one response shape, every product on the platform.
# Issue a token (one-time, from your account dashboard)
export ROSHAN_TOKEN="rk_live_..."
# Verify the token works
curl https://api.roshan-ai.com/v1/me \
-H "Authorization: Bearer $ROSHAN_TOKEN"curl -X POST https://api.roshan-ai.com/v1/shifamind/predict \
-H "Authorization: Bearer $ROSHAN_TOKEN" \
-H "Content-Type: application/json" \
-d '{
"note": "72M with HFrEF presents with progressive dyspnea, bilateral edema, BNP 1850. Started IV furosemide..."
}'{
"model": "shifamind-icd10-v2.1",
"codes": [
{
"code": "I50.23",
"description": "Acute on chronic systolic heart failure",
"confidence": 0.94,
"concepts": [
{ "label": "orthopnea", "activation": 0.93 },
{ "label": "lower_extremity_edema", "activation": 0.91 },
{ "label": "bnp_elevation", "activation": 0.90 }
],
"evidence": [
"progressive dyspnea",
"bilateral edema",
"BNP 1850"
],
"alternatives": [
{ "code": "I50.22", "description": "Chronic systolic heart failure" },
{ "code": "I50.9", "description": "Heart failure, unspecified" }
]
}
]
}Designed for production integration.
One response shape
Every product on the platform returns the same fields: predictions, confidence, concepts, evidence, alternatives. Integrate once; new products plug into the same handlers.
Auditable inference
Each response carries the activated concepts and verbatim evidence. The audit log is the response, not an addendum.
Stable contracts
Breaking changes ship behind versioned URL prefixes (v1, v2). Deprecations come with timelines, not surprises.
API access is gated while we onboard partners.
Tell us about your product, your data, and the workflow you're building. We work directly with integrators to size scope, price, and timeline before a key gets issued. Full developer docs land at docs.roshan-ai.com when the API moves to public access.