For organisations that already collect the data.

The forecast service

An application sends a meal and the person's recent glucose readings and receives a predicted glucose curve with an uncertainty band and an explanation of the contributing factors. Each request is self-contained, so partners do not need to move user histories into another system.

• Stateless forecasts

  Self-contained requests. The partner application decides what to show and when.

• Sensor push

  Glucose readings can be streamed as they arrive so forecasts start from the current state and can be scored against later readings.

• Accuracy reporting

  Every forecast is scored against the subsequently measured response. Reports are available per user and per organisation.

• Tenancy and data control

  Each organisation and its users are isolated. Usage is metered per tenant, and data erasure is available per tenant and per user.

Why a mechanistic model

A statistical model trained on user history can forecast the next hour of a familiar pattern. It cannot say what a different meal, a later dose or a walk would have done, because it has no mechanism to run forward under a change. SomaTapestry produces forecasts from a model of the physiology itself, so what-if questions can be answered and each forecast can be explained in terms of absorption, insulin response and the person's state at the time.

Research groups

Research teams can use the engine to test hypotheses against cohort data, with full provenance for every parameter and every simulation run. The integration roadmap is directed at the instruments research groups already use, including clinical-grade glucose monitoring, laboratory panels, metabolic and body-composition instruments, cardiac and sleep monitoring, and study data platforms. The platform page lists the planned integrations.

GLP-1 programmes

For programmes that prescribe or support GLP-1 medication, the model provides trajectory forecasts for weight and body composition, the fat-versus-lean split, withdrawal and regain dynamics, and the protein and resistance-training targets that protect lean mass. The platform supports the regimen set by the prescriber and does not make dosing recommendations. A clinician-facing export is planned.

Data protection and compliance

Data subject rights are built into the product: export in standard formats including a clinical interchange format, correction, consent versioning and deletion. Credentials are hashed, outbound notifications are signed, and the service is monitored. Formal security and health-data certifications are scheduled to begin alongside the first pilot partner.