Personal build · iOS 18.6+ · not on the App Store

One place for everything your body is doing.

Corpus pulls HealthKit, meals, peptide cycles, substances and lab reports into a single app — then gives an advisor the whole picture to answer from, instead of one slice of it.

Built by one person, for one person. There is nothing to buy and no waitlist to join. This page exists to describe what the app does.

Corpus / dashboard Read from HealthKit

HRV

62ms

Sleep

7:42

Resting HR

58bpm

VO₂ max

48ml/kg

Weight

82.4kg

Body fat

19.2%

Lean mass

66.6kg

Steps

9,140

In range Worth a look Out of range No threshold Thresholds vary by sex and age

01 — What it reads

Your devices already know. Corpus just asks.

Whoop, a smart scale and an Apple Watch all write to Apple Health. Corpus reads from there, so there are no per-device integrations to set up or break.

Body

Composition over time

Weight, body fat and lean mass, charted against your own history rather than a population average.

BMI and FFMI from your height
Sex- and age-aware thresholds
Colour-coded when a value drifts

Recovery

Sleep, HRV, resting heart rate

The signals that move first when something changes — a new compound, a bad week, a cut in alcohol.

Nightly sleep duration
HRV (SDNN)
Resting and respiratory rate

Output

What you actually did

Steps, active energy, exercise minutes and VO₂ max, so training load sits next to recovery instead of in another app.

Steps and active energy
Exercise minutes
VO₂ max trend

02 — What you log

The things no device can measure for you.

Four kinds of entry, each designed to take seconds. Everything is stored on the device first; syncing to your own server is what makes the history outlive HealthKit's window.

Meals

Photograph it, or just say it

A vision model reads the plate and estimates macros. Or describe the meal in words. Or scan a barcode and pull the real numbers from Open Food Facts.

Photo, text or barcode
Always editable before saving
Backdate a meal you forgot

Peptides

Cycles, doses and reminders

Start a cycle, log each dose with its route and amount, and get a local reminder when the next one is due. Sixteen common compounds come pre-loaded.

mg or mcg, six routes
Notifications that respect the cycle
Ends when you say it ends

Substances

Alcohol, nicotine, caffeine

The inputs that quietly explain a bad HRV week. Logged the same way as everything else, so the advisor can see them alongside the rest.

Amount, unit and route
Sits in the same timeline

Records

Blood work, read and filed

Upload a lab PDF or a photo of one. Pages are rasterised and passed to a vision model, which pulls out the individual findings and their values.

PDF or photo, up to 20 pages
Findings extracted separately
Available to the advisor

03 — The advisor

Answers that account for everything, not one slice.

Every chat starts with your recent biometrics, meals, active cycles, substance log and extracted lab findings already in context. It can also go and read the literature before answering.

Context

It already knows

No re-explaining what you are taking or how you slept. The advisor is handed the current state of every tracked signal before you type anything.

Literature

It can look things up

Search PubMed, ClinicalTrials.gov and Europe PMC preprints mid-answer, so a claim can be traced to a study rather than to a vibe.

Choice of model

Cloud, on-device, or your own box

Runs through OpenRouter by default. It can also run a local model on the phone, or reach an Ollama instance on your own hardware when a conversation should not leave the building.

04 — How it is built

Local first. The server is for memory, not for function.

The app works offline. Sign in with Apple mints a token for your own backend, which exists to keep history longer than HealthKit's rolling window — not to hold anything hostage.

  Whoop · Smart scale · Apple Watch
                 │
                 ▼
          Apple HealthKit          the source of truth for biometrics
                 │
                 ▼
    iOS app — SwiftUI + Spezi     meals, doses, records in SwiftData
                 │                    works with no network at all
                 │  HTTPS · Sign in with Apple · JWT
                 ▼
      FastAPI on a private VPS    long-term aggregation
                 │
                 ▼
   PostgreSQL 16 + pgvector  ·  Redis

iOS

Swift, Spezi, SwiftData

Built on Stanford's Spezi framework, which handles HealthKit access, onboarding and the LLM plumbing. Health data never touches a third-party analytics SDK, because there isn't one.

Backend

Python, FastAPI, Postgres

A small async service behind nginx, holding only what the app pushes to it: meals, cycles, substance entries and record extractions, each scoped to one account.

Honesty

What it will not do

Photo-based calorie estimates are a starting point and can be well off on portion size — the app asks you to confirm every one. Nothing here is medical advice, and the advisor reports your own logged data rather than prescribing from it.

05 — Status

Version 1, in active use.

This is a real app being used daily by the person who wrote it, which is why the roadmap is short and the finished list is specific.

ShippedHealthKit dashboard — eleven metrics, sex- and age-aware status, BMI and FFMI

ShippedMeal logging — photo, free text and barcode, with Open Food Facts lookup

ShippedPeptide cycles — dose log, routes, local reminders, sixteen seeded compounds

ShippedSubstance log and medical records with vision extraction of findings

ShippedAdvisor with full context and literature search across PubMed, ClinicalTrials.gov and Europe PMC

ShippedSync — Sign in with Apple, JWT, and a private backend for long-term history

BuildingDose schedules and titration — weekly cadences and escalation ladders, where the app asks before stepping a dose up rather than doing it silently

NextReconstitution calculator — vial, solvent and syringe units, because that arithmetic is where a tenfold error hides

NextBody measurements — tape-measure sites written back to Apple Health, and body fat estimated independently of the scale

NextTolerability log — side effects and injection sites, feeding the decision to hold or escalate