An independent, open-source iOS app that connects your QardioBase smart scale to Apple Health.
When Qardio Inc. shut down in 2025 — app delisted, servers dark — the scales kept working over Bluetooth. LibreBase is a small SwiftUI client that talks to the scale directly and logs your weight into HealthKit, so your data never depends on a vendor again.
Compatibility: tested only with the original QardioBase (1st generation, model B100). QardioBase 2 and QardioBase X are untested — they may work, but no promises. Have another Qardio Base? Please try LibreBase and tell us how it went — email librebaseapp@michelstorms.dev or send a capture log (see Contributing). Your reports help us support more scales.
- Direct Bluetooth LE connection to the QardioBase — no account, no cloud, no Qardio app required.
- Automatic weigh-in. Step on the scale; the reading appears and is saved to Apple Health.
- Apple Health integration. Weight is written to HealthKit's Body Mass.
- In-app BMI computed from your height (read from and synced back to Apple Health), with a unit-aware picker. The scale's own BMI is deliberately not used — see BMI & units.
- Locale-aware units. Weight shows kg or lb, height shows cm or ft/in, based on your region. Canonical metric values are always what's stored in Health.
- Battery and connection status at a glance, with a one-tap reconnect.
- Recon mode — an optional BLE capture log for inspecting the scale's GATT services and payloads, useful for supporting new hardware.
- iPhone running iOS 26.5 or later.
- A QardioBase (1st generation, model B100) smart scale — the only model tested so far. QardioBase 2 / X are untested; see Supported hardware.
- To build and run on a device: Xcode 26+ and an Apple Developer account (HealthKit requires a provisioned device, not the simulator, for real data).
git clone https://github.com/stormychel/LibreBase.git
cd LibreBase/iOS
open LibreBase.xcodeproj- Select the LibreBase scheme and your device.
- Set your own signing team (Signing & Capabilities → Team). The bundle id is
com.michelstorms.LibreBase; change it to your own. - Build and run. On first launch, grant the Bluetooth and Health permissions when prompted.
- Step on the scale.
The app is intentionally small — three types:
| File | Responsibility |
|---|---|
ScaleClient.swift |
CoreBluetooth: scans for the scale, connects, discovers services, parses weight, and exposes connection/battery state. |
Health.swift |
A thin HealthKit wrapper: authorization, saving weight, and reading/writing height. |
ContentView.swift |
The single SwiftUI screen: live reading, BMI, height picker, toggles, and recon log. |
The QardioBase is discovered by name and advertised service, then connected. LibreBase reads the weight from whichever of two paths the unit exposes:
- Standard BLE Weight Measurement (
0x2A9D) — the SIG-spec'd payload, normalized to kilograms (handling the imperial flag). - Qardio proprietary result — after the scale signals "done", LibreBase reads characteristic
B24F98BE-…, which returns a plain-JSON final measurement.
In both cases weight is normalized to kilograms internally and only converted for display.
BMI is computed in-app, never taken from the scale. The scale can only report BMI if it knows your height, and that height could previously only be set through the now-discontinued Qardio app — so for most users it is stale, belongs to another person, or is absent entirely. Instead, LibreBase:
- reads your height from Apple Health to seed the value automatically,
- lets you adjust it with a wheel picker (cm, or ft/in by locale), and
- writes any edit back to Health, so the two never diverge.
BMI is then weight / height², shown only once a height is known.
Toggling Recon mode makes the client discover every service and characteristic and log each payload as hex. This is how the protocol below was mapped, and it's the starting point for adding support for other models.
LibreBase has no servers, no analytics, and no network code. Everything happens on-device over Bluetooth and through HealthKit. Your weight and height never leave your phone except where you sync HealthKit via iCloud. The app declares ITSAppUsesNonExemptEncryption = false (it uses no encryption beyond the OS).
| Model | Status |
|---|---|
| QardioBase (1st gen, B100) | ✅ Tested — weight → Health works |
| QardioBase 2 | ❓ Untested — likely similar profile |
| QardioBase X | ❓ Untested — newer/rechargeable, may differ |
Body composition (fat %, water, muscle, bone) is present in the scale's JSON but not yet parsed or saved — see the roadmap.
- Parse and save body composition (bioimpedance → fat %, lean mass) to HealthKit.
- Verify and broaden QardioBase 2 / X support.
- Weigh-in history view in-app.
- Contribute a QardioBase driver to openScale on the Android side.
Issues and PRs welcome — especially capture logs from a QardioBase 2 or X (use Recon mode and paste the log). If you're adding a model or decoding body composition, the protocol notes below are the place to start.
MIT © Michel Storms. Note: openScale is GPL — treat its impedance formulas as a reference, not copy-paste, in this permissively-licensed app.
- LibreArm — the open-source QardioArm revival that proved the approach (BLE reverse-engineering + HealthKit, approved on the App Store as a wellness app).
- openScale (oliexdev) — the reference open-source scale project and its "how to support a new scale" workflow.
- Nordic nRF Connect / nRF Sniffer and Apple PacketLogger — the BLE recon and capture tooling.
These are the findings behind the implementation, kept for anyone supporting new hardware or decoding body composition.
The tested unit does not expose the standard Weight Scale (0x181D) or Body Composition (0x181B) services. It exposes Device Info, Battery, and a custom Qardio service:
| UUID | Meaning observed |
|---|---|
A78AF805-8F3F-4E8F-A964-318B768BC38C |
control/state notify: 00 idle, 03 measuring, 06 done |
9F3F4E1B-37D7-4F95-B374-CF585D808BEB |
noisy engineering/debug stream; also accepts config write 00 00 01 01 |
B24F98BE-9CD4-4F82-B935-01F18F104EDE |
final measurement JSON; read after state 06 |
1EC92A15-14E0-43E7-A990-CB37000990BA |
calibration JSON |
After state 06, reading B24F98BE-… returns plain UTF-8 JSON. This is the reliable source LibreBase uses (only weight is consumed today):
{"weight":"76.0","bmi":"19.3","z":"2031","fat":"57","tbw":"31","bmc":"3","mt":"9","sm":"17"}fat, tbw (total body water), bmc (bone mineral content), mt, sm (skeletal muscle) are the body-composition fields awaiting a parser. The scale-provided bmi is ignored (see BMI & units).
You do not need this if you read the JSON, but for reference: the 9F3F… stream carries raw load-cell frames, and 1EC9… exposes calibration:
{"0":"0","50":"5945","100":"11888","150":"17836","z":"523"}That's ≈ 118.907 raw counts per kg, with z a millikilogram offset (523 → +0.523 kg). A known-good frame for an expected 87.4–87.7 kg weigh-in:
21 70 28 10
raw = littleEndian(bytes[1...2]) = 0x2870 = 10352
kg = raw / 118.907 + 0.523 = 87.6 kg
Many of the engineering stream's 16-bit windows are false positives and can repeat between users — prefer the JSON characteristic.
- Enable Recon mode in the app and step on the scale — the log shows every service, characteristic, and payload.
- For deeper captures, an nRF52840 dongle + nRF Sniffer (into Wireshark, filter
btatt) records the live link with best fidelity. - Note your exact model and serial; QardioBase 2 / X may differ in services or firmware.
Reverse-engineering hardware you own from a defunct vendor for interoperability is broadly defensible, but this is engineering documentation, not legal advice.