An open standard so individuals own and port their health & fitness data — without platform lock-in. A canonical data model plus mapping semantics, implementable in any language.
OpenBody is an open, vendor-neutral, language-agnostic standard: a canonical data
model plus mapping semantics that anyone can implement in any language. Its mission is
that individuals own and port their health and fitness data without platform lock-in;
its adoption lever is the apps and platforms that implement it to exchange that data
losslessly.
It is fully open, not open-core — the entire model and its mapping rules are public.
The model is defined independently of any serialization; JSON (with a published JSON
Schema) is the primary, first-class binding. OpenBody is stewarded by Thabit Labs,
with intent to transfer to a neutral foundation as adoption warrants.
The two pillars
Pillar A — Observation (telemetry)
Continuous or instantaneous measurements: heart rate, HRV, sleep, vitals, power,
speed, cadence, GPS. Readiness and recovery are observations, not training attributes.
Defined by the canonical Measurement (SPEC §4).
Pillar B — Structured Training (exercise)
The prescription and execution of training — programs, sessions, blocks,
exercises, work units, reps, with targets, effort, equipment, and the
planned-versus-performed relationship. This is what no incumbent provides cleanly
(SPEC §5–§6).
Both pillars are first-class. A structured-training record never re-invents telemetry:
Pillar B references Pillar A Measurements through typed links (§1.3, §7). Every
record — in either pillar — shares a common record envelope (§7): a stable identifier,
typed links, provenance, and an immutable-plus-supersession lifecycle.
Why it exists
Personal health and fitness data is trapped in silos. Every platform — Apple HealthKit,
Google Health Connect, Garmin, Fitbit, Oura, Whoop, and a long tail of strength, endurance,
and functional-fitness apps — defines its own schema, units, identifiers, and aggregation
rules. There is no neutral, portable representation all of them can read and write.
Telemetry is partially standardized; structured training is barely served at all.
OpenBody is an open standard for both.
Start here
Getting startedInstall openbody-ts, validate and normalize a record, run the vectors.
ConceptsThe data model, the two pillars, exercise identity, and §8.3 canonicalization — explained plainly.
SpecificationThe rendered, canonical SPEC.md (v0.3.1) plus the JSON Schema.
ConformanceProfiles, the test vectors, and how to claim “OpenBody Compatible”.
RegistryThe hand-curated canonical exercise registry, facets, and crosswalks.
“OpenBody” is a vendor-neutral standard stewarded by Thabit Labs. The steward is not the
brand on the wire. Specification: OWFa 1.0 · Reference code:
Apache-2.0 · Registry data: CC0. See
Licensing.