Mirobody

Mirror Your Body in Data

One standard, every source, one complete picture of you — building a lasting record to understand where your health is going. This is Mirobody.

Try Mirobody in Theta Web →
Open infrastructure Proven in production
1,400+ GitHub stars #1 digital-health · #2 health-data 30,000+ Users 800+ Daily active Powering Theta app

Health data engine

600,000 raw names,
1,000 standard indicators

The health-data engine behind Theta: every reading from labs, wearables, records and manual logs maps to one vocabulary and unit—comparable, computable, traceable.

600,000+
Raw indicator names
1,000+
Standard indicators
LOINC · ICPC-3 · UCUM
Lab, clinical & unit standards

In practice

One indicator, dozens of names.

  • LDL
  • LDL-C
  • Colesterol LDL
  • LDL Chol
  • ldlCholesterol
ldl_cholesterol LDL Cholesterol Unit · range · conversion rules

Apple Health · dailyTotalSleepTime

27,081,674 msSleep time 451.36 min

Lab report · Glucose

6.1 mmol/LGlucose 110 mg/dL

Low-confidence mappings go to the user.

Indicator system

Define once, use everywhere.

1,000+

Indicator catalog

Canonical names, units, reference ranges and conversion rules, stored once as the system’s single source of truth.

By condition

Condition knowledge base

Diabetes, hypertension, lipids, CKD, heart failure, osteoporosis and more: indicators, alerts, complications, screening.

Below ↓

Advanced indicators & models

Computes Body Age, Fracture Risk and more, and runs our in-house age / disease / mortality prediction model.

Recognized health data types

All figures approximate
Mirobody 1,000+
Apple Health ≈190
Google Health Connect ≈40

Linear scale. Apple Health ≈ 190 HealthKit sample types (120 quantitative + 70 category, iOS 26); Google Health Connect ≈ 40 record types across 7 categories. The point is the order of magnitude, not the decimal.

Standardization

Six steps, each on record.

  1. 01 · Auto

    Ingest

    Labs, records, wearables, genomics, or numbers in chat.

  2. 02 · AI

    Mapping

    Each reading → standard name, ID, unit, category.

  3. 03 · Auto

    Units

    Compatible units convert: glucose 6.1 mmol/L = 110 mg/dL.

  4. 04 · Doctor

    Review

    Tag, merge duplicates; fix names, units, ranges.

  5. 05 · User

    Confirm

    Unsure matches go to the user; reused on later imports.

  6. 06 · Auto

    Provenance

    Every cited reading traces to its source report.

Quality gateZero misses, false hits and mix-ups to merge; agents propose, humans decide.

Where your Mirobody lives

One record, kept where you choose.

Kept by you

Open-source edition

On your own infrastructure: reports, wearable data and genetic files in a standardized record, with source tracing and MCP access for your AI agent.

Self-host guide → GitHub

Standardized records · illustrative examples

Names, units and categories — illustrative examples
std_idCanonical nameUnitSystemKind
ldl_cholesterolLDL Cholesterolmg/dLlipidlab
hemoglobin_a1cHemoglobin A1c%glycemiclab
resting_heart_rateResting Heart Ratebpmcardiovascularlifestyle
respiratory_rateRespiratory Ratebreaths/minrespiratorylifestyle
sleep_efficiencySleep Efficiency%lifestylelifestyle
phenotypic_agePhenotypic Ageyearsderivedderived

Human deep data

Standardize enough of your data, for long enough, and something new appears: a complete, continuous account of one human body. We call it your Mirobody.

The health equivalent of a company's financial statements — not a snapshot from a single visit, but a longitudinal ledger deep enough to reason about, and in time, to forecast from. This is the asset that compounds the longer you hold it.

Depth of record · over time

One body, accumulating toward a model of itself.

First importYears of record

Advanced indicators

What no single reading shows.

Mirobody computes these from the standardized record: patterns that no single measurement shows on its own.

  • Body Age

    yrs

    Biological age from 9 routine labs + age.

  • Heart Age

    yrs

    10-year heart risk, expressed as an age.

  • Body Composition

    Fat, waist, lean mass.

  • Cardio Fitness

    VO₂max percentile vs. same age and sex.

  • Sleep Score

    Duration, wake, onset, HRV, RHR; 0–100.

  • Stress Score

    Autonomic load from HRV; 0–100.

  • Fracture Risk

    %

    5- and 10-year fracture risk, for older adults.

  • Clinical

    Guideline

    10-yr CV risk (clinician-side), eGFR.

These advanced indicators are available in Theta ↗. Theta's app and web offer advanced indicators and broader standardization through internal, closed-source Mirobody capabilities, beyond the open-source edition.

Advanced model · In-house

Age, disease and mortality prediction.

12-layer Transformer encoder, ~87M parameters. Takes ~800 lab and exam items in any number or order; outputs biological age, 5/10-yr mortality, current risk of 29 diseases and 10-yr onset risk of 28.

Biological age · MAE in healthy adults

Our model PhenoAge yrs

Group A Our model 1.73 PhenoAge 7.47
Group B Our model 3.39 PhenoAge 8.55
Group C Our model 3.86 PhenoAge 6.84

Same healthy people: error significantly below PhenoAge, GOLD BioAge and Bortz BAA. Individual error is still notable; watch trends.

Disease & mortality · discrimination

Our model Age + sex only AUC

Mortality · 10-yr Our model 0.935 / Age + sex only 0.876
Diabetes · current Our model 0.943 / Age + sex only 0.654
Diabetes · 10-yr Our model 0.901 / Age + sex only 0.633
Mortality · CBC only Our model 0.901 / Age + sex only 0.872

Independent test set, evaluated once; AUC 0.5 = chance.

No guessing: lists missing inputs, never invents numbers; for health guidance, not diagnosis.