Export

Everything you have logged, or just a window of it, as one JSON document. It is built on the phone and handed to the iOS share sheet; it goes only where you send it. Field names read as English and dates are ISO 8601, so the file works as direct input to a spreadsheet, a script or a language model without a decoder ring.

Settings: the data server field, the installed bundle version with Check for Updates, and the Your data section with a timeframe picker, Include Apple Health workouts and Build a JSON export.

Making one

Settings › Your data. Pick a timeframe, tap Build a JSON export, then Save or share the export. The summary under the button leads with the span the file covers and counts the diary entries, workouts, Apple Health workouts and habit ticks inside it, so "I thought I exported just yesterday" is caught before sharing. Changing the timeframe discards a built file; the share button never hands out a file from a different span.

Timeframe	Covers
Today	Today only.
Yesterday	Yesterday only.
Last 7 days	Today and the six before it.
Last 30 days	Today and the 29 before it.
Last 90 days	Today and the 89 before it.
Custom range	Two dates you pick, inclusive, neither in the future.
Everything	Full history.

Windows are whole calendar days in your own time zone, the same grouping the diary uses. The file is named for what it holds: setpoint-export-2026-10-11-to-2026-10-17.json for a window, setpoint-export-2026-10-17.json for full history, dated the day it was built.

How to read the file

schema is 8. It rises when the shape changes.
covers gives the first and last day of a window. When it is absent the file is full history. When it is present, a reader must not treat a missing day as an unlogged day.
timezone is the IANA zone every timestamp is expressed in, such as America/New_York, so a tool groups days the way you do instead of guessing.
Timestamps carry your offset, as in 2026-10-17T22:14:00-04:00, rather than UTC. A ten p.m. dinner belongs to that calendar day and any tool grouping by date must see it there. Still RFC 3339; every standard parser reads it. Day fields such as day, firstDay and healthWriteSince are plain dates, 2026-10-17.
Numbers have at most two decimals. Finer than anything the app measures, and it keeps float noise out of a diff between two exports.
Keys are sorted and the file is pretty-printed, so two exports diff cleanly.
A window trims the time series only. Diary entries, weigh-ins, composition, measurements, workouts, done marks and habit ticks are filtered to the window. Library data always ships: profile, current targets, templates, exercise settings, recipes, custom foods, custom exercises and custom habits, because a subset is unreadable without its definitions. The composition summary is also always present and judged on the whole history.
Absent means unknown. Optional fields are left out rather than written as zero or null. Fiber that was never known is simply not there.

What is in it

Key	What it holds
schema, exportedAt, timezone, appVersion	The file's own facts.
covers	`firstDay` and `lastDay` of the window. Absent for full history.
profile	Date of birth, sex, height, weight, activity level, weight goal, pace in kg per week, target weight and date if set, protein mode and figures, unit preference, and the goal metric (`weight`, `bodyFatPercent` or `fatMass`) with its target and the date it last changed. The goal metric is never an input to the targets.
currentTargets	Today's typical budget: kcal, protein, carbs, fat and fiber in grams, computed from the profile and the baseline movement figure.
dailyTargets	One row per logged day: the day, Apple Health's measured movement for it when there was any, the `basis` the budget used (`measured`, `projected`, `estimated`, `coldStart`) and the targets the app displayed that day. Without these a tool comparing entries against `currentTargets` concludes the diary is missing food that was never missing. Judged by the current profile, like everything else.
foodDiary	Each entry, oldest first: `loggedAt`, `meal`, `name`, `brand`, `quantity`, `portion`, `grams`, `kcal`, `proteinG`, `carbG`, `fatG`, `fiberG` when known, `source` (the database it came from), `weightIsNominal` when the grams are a stand-in for a counted serving rather than a weight, and `uuid`, the same id the entry's Apple Health samples carry when the food log is shared.
weighIns	`date`, `weightKg`, `fromHealthKit`, and `source`: `manual` for a typed weigh-in, otherwise the Health source's name, so a reader can tell your number from the scale's.
bodyComposition	One row per day a scale reported: `date`, `bodyFatPercent` (a percent, 28.4, not a fraction), `weightKg` (the weight the masses were derived from: the scan's own, else the day's weigh-in), `fatMassKg` and `leanBodyMassKg` (weight × percent, and the rest, so fat + lean = weight), `reportedLeanBodyMassKg` (what the scale said, kept for the record), `coherenceFlag` (present, and true, only when the reported lean mass disagrees with the derived one by more than 50 g), `source`. Masses are absent when no weight existed to derive them. Whatever `covers` says, every reading in the median's 14-day window is included, so the headline can be checked against the rows it came from.
currentBodyComposition	Where composition stands as of the export, from the whole history, whatever the window: the `latest` reading; `median`, the median reading of the last 14 days as one full row (date, bodyFatPercent, fatMassKg, leanBodyMassKg, weightKg, source), the readings sorted by body fat percent and the lower middle one taken, so it is a real measurement whose figures balance; `readingsInWindow` and `medianWindowDays` (14); the change in body fat, fat mass and lean mass against the reading nearest 30 days back; and `stale`, true when the newest reading is over 14 days old. `median` is absent, not zero, under three readings in the window or when stale, and the trend then stops at `latest`.
healthWriteEnabled, healthWriteSince	Whether the food log is being pushed to Apple Health and from which day, so a reader knows another store may hold a copy of the diary. `healthWriteSince` is always a real day: when the whole log is shared it is the first diary entry's day. Present only while sharing is on.
measurements	Measured days: `date`, `note`, `sites` (each with `site`, `centimeters`, `inches`, `fromHealthKit`) and `photos` (each with `filename`, `pose`, `date`). Photo bytes are not in the file.
workouts	Everything trained, in one list, oldest first, the way Session history shows it. `source` is `logged` for a session with sets entered here or `apple-health` for one only the Watch knows about. A logged session carries its `plan` (the prescribed sets with their progression decision record), its `sets`, its `notes`, an optional `watchRecording` (the Health record of the same workout: activity, duration, kcal, average heart rate, and which app recorded it), and `lightSession` when it was planned light. One workout, two records of it, never two workouts.
unattachedSets	Sets whose session is not in the file.
workoutTemplates	Each template's name and exercises in order.
exerciseSettings	The progression rules per exercise, as suggestions applied them: increment, equipment, floor and bar weight, band levels, rep range, set structure, working and warm-up set counts, gate, maximum jump, technique and retired flags, calibrating state.
exercisesMarkedDone	Exercise and day for each "done" mark.
customExercises	Name, primary muscle, equipment, notes.
habitChecks	Each tick: `habit`, `day`, and `origin`, which is `auto` for a tick the app credited and otherwise a manual one.
activeHabits	The habits currently chosen, by slug.
customHabits	Name and cue.
recipes	Name, servings, source link, instructions, the original ingredient text, and each ingredient with its matched name, quantity, portion, grams, `weightIsNominal` when the grams are a stand-in, kcal, protein and fiber.
customFoods	Name, brand, values per 100 g, or `perServing` naming the serving the values are for.

Weights in the training log

Every set and planned set carries weightKg and weightLb. A lifter reading plan against actual thinks in plates, and 34.02 kg hides that it was 75 lb. Sets also carry setNumber, reps, completedAt, and when present rating, failReason, deviation, pain, side and warmup. Planned sets carry the decision record: rule, baseLb, deltaIncrements, previousLb, previousReps, previousRating, gateRemaining, ladderRepaired, governingSet and explanation. The rules are described on the Train page.

Deliberately absent

Caches, day locks, and the per-food correction plumbing (custom portions and nutrition overrides). Diary entries snapshot their nutrition at logging time, so the record is complete without the machinery behind it. Photo bytes, which live in the app's folder. Anything from Apple Health that you did not connect.

An abridged example

A one-day file, cut down to show the shape. Values are illustrative.

{
  "appVersion" : "1.0",
  "covers" : {
    "firstDay" : "2026-10-17",
    "lastDay" : "2026-10-17"
  },
  "currentBodyComposition" : {
    "bodyFatPercentChange30Days" : -0.8,
    "latest" : {
      "bodyFatPercent" : 24.1,
      "coherenceFlag" : true,
      "date" : "2026-10-17T06:42:00-04:00",
      "fatMassKg" : 19.52,
      "leanBodyMassKg" : 61.48,
      "reportedLeanBodyMassKg" : 61.4,
      "source" : "Withings",
      "weightKg" : 81
    },
    "median" : {
      "bodyFatPercent" : 24.4,
      "date" : "2026-10-12T06:40:00-04:00",
      "fatMassKg" : 19.86,
      "leanBodyMassKg" : 61.54,
      "source" : "Withings",
      "weightKg" : 81.4
    },
    "medianWindowDays" : 14,
    "readingsInWindow" : 6,
    "stale" : false
  },
  "dailyTargets" : [
    {
      "basis" : "projected",
      "day" : "2026-10-17",
      "movementKcal" : 412,
      "targets" : { "carbG" : 225, "fatG" : 47, "fiberG" : 20, "kcal" : 1703, "proteinG" : 96 }
    }
  ],
  "exportedAt" : "2026-10-17T21:05:12-04:00",
  "foodDiary" : [
    {
      "carbG" : 1.1,
      "fatG" : 9.5,
      "grams" : 100,
      "kcal" : 143,
      "loggedAt" : "2026-10-17T08:05:00-04:00",
      "meal" : "breakfast",
      "name" : "Egg, whole, raw, fresh",
      "portion" : "2 large",
      "proteinG" : 12.6,
      "quantity" : 1,
      "source" : "usda_foundation",
      "uuid" : "6B1C2F0E-3D7A-4C7B-9E1F-2A0B5C8D9E10"
    }
  ],
  "healthWriteEnabled" : false,
  "schema" : 8,
  "timezone" : "America/New_York",
  "weighIns" : [
    {
      "date" : "2026-10-17T06:41:00-04:00",
      "fromHealthKit" : true,
      "source" : "Withings",
      "weightKg" : 81
    }
  ]
}