Raw signal, in the hands of developers and institutions.
From band to cloud, every layer is designed for developers and agents: an SDK that gets you raw signal, cloud services for sync and analytics, and interfaces written for machines first.
iOS and Android, raw signal straight to you.
Public beta
Raw capture
Raw PPG and acceleration streamed in real time; standalone acceleration capture for up to 7 days.
Typed history
15 types of device history—heart rate, sleep, SpO2, HRV and more—output by type, with resumable incremental sync.
On-demand measurement
Trigger heart-rate, SpO2 and HRV measurements and read each result.
Resumable, exportable sessions
Capture sessions persist and resume after interruption, with typed export under the data contract.
Build without a band
Replay recorded real data; self-check tools help diagnose connection and sync issues.
Local mode
Use the SDK alone and data never touches our cloud.
iOS and Android binaries · evaluation licenceContact us →
Data sync, gap filling and analytics.
In development
One account, many bands
Data is attributed by binding period, so swapping bands or wearing two stays clean.
Algorithms in the cloud
Sleep, recovery and load scores run in the cloud and improve as the model does; the Soma-ACC activity-recognition model is already serving inference there.
Interfaces for agents
Data contracts, OpenAPI and /llms.txt; MCP is already running in a test environment.
Hosted and authorised
Hosted first: the SDK plus our cloud. Authorised later: users of our app grant access to third parties.
Data stays in region
China and international data are stored separately.
We are the first tenant
SomaLoop for BioHacker runs on these same cloud services.
Ecosystem and open-source releases
SomaLoopMotion
A tool for aligning video with band acceleration, segmenting events and clustering motions, developed by a team at Central South University. The code is public on GitHub.
github.com/NKWBTB/SomaLoopMotion →Soma-ACC, open-source model
The first Soma model: wrist-accelerometer activity recognition, coming soon on HuggingFace under Apache-2.0.
About Soma-ACC →Discipline across every interface
- 0 = not measured, never reported.
- Quality flags travel with the sample; failed samples are stored but never interpreted.
- The wrist temperature field is skin_temperature and never impersonates body temperature.
- Other brands' proprietary scores are quarantined and never enter standard metrics.
- The language model never generates a physiological number.