Built like a clinical system, not a chatbot.
Every piece of patient data lives behind row-level security. Every model decision is logged. Every release passes a clinical governance review.
What we do, and what we publish.
The guardrails.
Row-level security.
Every row a clinician reads, every message a patient writes — guarded by Postgres RLS policies we publish and a governance committee audits.
MFA for every clinician.
TOTP enrollment is mandatory for clinician, supervisor, and admin roles. Enforced in middleware and in the database itself.
Adults 18+, outpatient only.
Scope discipline is a safety feature. We do not serve minors and we do not serve inpatient crisis — by design, not oversight.
Audit log you can read.
Every model response, every escalation, every data access is append-only logged with an actor, a reason, and a timestamp.
Encryption in flight and at rest.
TLS 1.3 in transit. AES-256 at rest. Field-level encryption for the most sensitive clinical attributes, with key rotation on a quarterly cadence.
Clinician-supervised model updates.
No model update reaches patients until a clinical governance committee has reviewed the canary evaluation and signed off in writing.
Every access, every model decision, every escalation.
Queryable by patient, actor, action, date — exportable to your SIEM on System tier.