The care platform where AI drafts and humans decide
Live demo · one click into any role · synthetic data only
0
Governed agents
Every one proposes; humans decide
0
Care settings
AL, BH, IDD, group homes, home care
0+
Automated tests
Tenant isolation tested adversarially
0
Unaudited sensitive actions
Denied attempts are audited too
Real platform figures, not marketing rounding. ClinivX is a prototype running synthetic data only.
The governed pipeline
Draft → Validate → Approve → Record. Every agent, every time.
There is no path where a model writes to the clinical record. The pipeline below is the only way AI output becomes a record, and deterministic code enforces every step of it.
01 · Draft
Draft
An agent reads only what its governed scope allows and drafts: a visit note from session facts, an MDS Section GG code from the look-back window, a follow-up for a critical result. Deterministic code gathers every fact; a model only words the narrative.
02 · Validate
Validate
Every model-written draft passes a deterministic evaluation gate: grounded in the supplied facts, free of injection markers, inside its length bound. A failed check flags the draft for review; it never reaches the record silently.
03 · Approve
Approve
A licensed human with the same domain permission the manual action needs reviews the proposal in the AI inbox. Approval is recorded and IS the signature; rejection and dismissal are recorded too. The AI can never exceed the approver's own rights.
04 · Record
Record
Execution runs through the same domain service a person would use: every permission, integrity rule, and append-only constraint re-applies, and the action lands in the hash-chained audit trail with the human as the actor.
A day inside ClinivX
Built for the people running the building
The look-back window, already assembled
- 1The MDS Agent aggregates every note, lab, and incident in the look-back window with source links
- 2Section GG drafts arrive coded by the deterministic coder, with a Green/amber/red confidence review
- 3You reconcile with clinical judgment, approve, and lock. Your lock is the signature
Nothing auto-submits, and no code exists that you didn't approve.
Every module, one record
Stop stitching point tools together
Resident records
One governed profile per resident, facility-scoped by construction.
Progress notes
Append-only versions; signing freezes; addenda carry reasons.
Care plans
Versioned plans with four-eyes approval and goal-to-note links.
eMAR
Pass windows, controlled-substance witnessing, barcode evidence.
Incidents
Immutable versions, per-type regulatory deadlines, state packs.
Scheduling
Tasks with completion evidence; shifts with coverage detection.
MDS
Look-back aggregation, deterministic Section GG coding, your lock.
Therapy (PT/OT/ST)
Plans of care, GG-tied goals, self-computing recert windows.
Pathology
Append-only reports with an acknowledgement trail and escalation.
Analytics
Aggregate-only dashboards over the live record. PHI-safe.
Survey readiness
The binder, credential alerts, and QAPI trends, always current.
Family portal
Guardian access that is the authorization, not a copy of it.
What operators tell us they need
Illustrative composite scenarios from operator conversations, not customer endorsements. ClinivX is a prototype.
“The survey binder used to be a weekend. Now it's a button, and every number in it traces to a record.”
Weighing your options?
Honest comparisons against the categories operators actually shop: legacy EHRs, therapy point tools, and BI bolt-ons.
See ClinivX run a real day
One click into any role: the nurse's med pass, the coordinator's MDS queue, the therapist's plan of care, the surveyor's binder. Synthetic data, real governance.
ClinivX is a prototype demonstrating these controls; it is not a HIPAA-compliance certification and holds no real protected health information.