Platform · MSI CORE

Medical Super Intelligence. The AI Agent Layer.

MSI CORE is cognitive orchestration: agents that decompose a task, call the right domain product, assemble the answer and present it with its evidence — under governance that decides what they are permitted to do.

Superhuman coverage. Human authority.
Where it sits

Between the data fabric and the domain products.

MSI CORE does the intelligent work. The Trust Fabric governs how that intelligence is allowed to work. They are separate on purpose — an engine that also writes its own rules cannot be audited against them.

01
Healthcare Data Fabric

Supplies governed, resolved, interoperable data.

02
MSI CORE — AI Agent Layer

Orchestrates. Plans, retrieves, routes, composes, presents.

03
OmnigeniX

Supplies and interprets multi-omics intelligence.

04
CytoMind

Creates and operates the Digital Health Twin.

05
TheraMind

Applies intelligence to care orchestration.

06
Applications

Clinical, research, consumer, government, partner and programme surfaces.

Soma Nova = Healthcare Data + MSI CORE AI Agents + Domain Intelligence + Responsible AI Governance.

The agent families

Six families, and the orchestration between them.

Each agent has a declared task, a declared data scope and a declared output type. A general-purpose assistant has unbounded scope and therefore cannot be validated, registered or defended.

Clinical AI Agents

Prepare the consultation — history, trajectory, interactions and candidate actions — for review and signature.

Genomics AI Agents

Assemble the molecular picture, retrieve the evidence, and draft over cited findings only.

Preventive AI Agents

Risk trajectory, screening prioritisation and early-signal detection ahead of clinical onset.

Care AI Agents

Pathway execution, follow-up, adherence and escalation across the care team.

Research AI Agents

Cohort discovery and multi-omic analysis, executed federated against custodian data.

Population Health Agents

Stratification, programme analytics and equity reporting at cohort scale.

Orchestration

Task decomposition, agent routing, state handling and composition — the layer that decides which agents are involved in a request and in what order, and records that decision as part of the provenance record.

The boundary

What an agent may and may not originate.

01
Agents plan and route

Decompose a task, select the domain product, sequence the calls.

02
Agents retrieve

Fetch evidence through governed access, never by direct reach into a store.

03
Agents compose and present

Assemble findings into the shape of the work a person is doing.

04
Agents draft

Narrative over cited findings. Clinical numerics are rendered from the structured record by substitution, never generated.

05
Agents do not classify

Classification runs on OmnigeniX versioned, inspectable criteria rules.

06
Agents do not decide

Every clinically consequential output passes the sign-off gate and is signed by a credentialed person.

07
Agents are registered

Each carries an AI registry entry and an agent passport: task, permitted data, output type, risk class, evaluation record and approval state.

Medical Super Intelligence

Superhuman coverage. Human authority.

The claim is specific. No clinician can hold four million variants, the current literature, the interaction surface and a decade of longitudinal state at once. MSI CORE can. That is superhuman in coverage, consistency and recall — and it holds no authority at all.

ClaimedSynthesis across a breadth of evidence and a length of history no individual can hold simultaneously.
ClaimedConsistency across cases, sites and time; continuous re-interpretation as evidence changes.
Not claimedAutonomous clinical judgement, diagnosis without a clinician, or independent decision authority.
Not claimedLearning in production without an approved change control plan.
Not claimedReasoning that cannot be inspected.
Get started

See the agents on your workflow.

Anchor sites are partners, not customers — early capability and a real say in the roadmap.