Bissbilanz

Know every bite. Own every goal.

A calorie and macro tracker for web, Android and iPhone — barcode scanning, on-device nutrition-label OCR, offline-first sync, and a food log you can hand to an AI agent.

Open the web app · Join the iOS beta · Join the Android beta

---

Try it

🌐 Web	bissbilanz.orellbuehler.ch — installable PWA, works offline
🍎 iPhone / Apple Watch	TestFlight beta — widgets, watch app, Apple Health, fasting Live Activity
🤖 Android / Wear OS	Google Play beta — opt in, then install from Play; Wear OS app, Health Connect, widgets

Free, no ads, no tracking SDKs, no data selling. Try the mobile apps without an account at all — local-only mode keeps everything on device and migrates into your account if you later sign in.

What makes it different

Your food log is an MCP server. Point Claude (or any MCP client) at it and just say what you ate. 63 tools cover logging, foods, recipes, goals, weight, sleep, supplements, analytics and the AI task queue — OAuth-protected, so the agent only ever sees your data.

On-device label OCR. No barcode? Point the camera at the nutrition table. A shared Kotlin parser plus ML Kit reads the values locally — nothing leaves the phone.

Offline-first, for real. Every client writes optimistically to a local store (Dexie on web, SQLDelight on Android, SwiftData on iOS) and drains a sync queue with idempotency keys and last-write-wins conflict resolution. Log on a plane; it reconciles when you land.

43 extended nutrients. Beyond calories and the five macros — vitamins, minerals, amino acids — sourced from Open Food Facts and your own database.

Actually native. Not a wrapped web view: Jetpack Compose on Android, SwiftUI on iPhone, plus a Wear OS app, an Apple Watch app, home- and lock-screen widgets, and Health Connect / Apple Health integration.

Features

Track	Calories, protein, carbs, fat, fiber + 43 extended nutrients, per meal and per day
Log fast	Barcode scanner, camera label OCR, food photos, favorites, recent foods, one-tap widgets
Recipes	Multi-ingredient recipes with automatic per-serving nutrition
Beyond food	Weight trend, sleep, supplements, fasting timer with Live Activity
Insights	Maintenance-calorie estimate from weight trend + intake, streaks, meal timing, food diversity, sleep/food correlation
AI	Natural-language logging via MCP, an agent task queue in both mobile apps, on-device meal estimation from a photo on iOS
Sync	Web, Android, iOS and watch stay in sync; conflict-safe and offline-tolerant
Accounts	Infomaniak, Google or Apple sign-in — or no account at all on mobile
Languages	English and German

How it fits together

flowchart LR
    API["SvelteKit API on Bun<br/>Zod-validated · OpenAPI-generated clients"]
    DB[("PostgreSQL<br/>Drizzle ORM")]
    OFF["Open Food Facts"]

    subgraph Clients
        direction TB
        W["Web PWA<br/>Svelte 5 + Dexie"]
        A["Android + Wear OS<br/>Compose + SQLDelight"]
        I["iPhone + Watch<br/>SwiftUI + SwiftData"]
    end

    C["Claude / any MCP client"]

    W -- "REST + sync queue" --> API
    A -- "REST + sync queue" --> API
    I -- "REST + sync queue" --> API
    C -- "/api/mcp (OAuth)" --> API
    API --> DB
    API --> OFF

Loading

The Zod validation schemas are the single source of truth: bun run api:generate emits docs/openapi.json and from it both the TypeScript and Kotlin client models, so the web, Android and iOS clients can't silently drift from the server.

Talk to your food log

Bissbilanz exposes a remote MCP server at /api/mcp (streamable HTTP; OAuth 2.1 with PKCE; clients are provisioned in Settings → MCP, scope mcp:access):

{
	"mcpServers": {
		"bissbilanz": {
			"type": "http",
			"url": "https://bissbilanz.orellbuehler.ch/api/mcp"
		}
	}
}

Then:

You: I had a chicken bowl with rice and avocado for lunch, and a flat white.

Claude: Logged 4 items to Lunch — 812 kcal, 47 g protein. You're at 1,340 / 2,200 kcal and 89 / 160 g protein for the day.

Anything the app can do, the agent can do: log_food, search_foods, create_recipe, get_daily_status, log_weight, get_streaks, get_sleep_food_correlation and 56 more — see docs/mcp.md for the full tool, prompt and resource list and how to connect each client.

An agent can also label the food database — list_unlabeled_foods plus set_food_labels gives every food the plain English noun a camera would call it, which is how the phone matches a food from a camera frame.

Tech stack

Layer	Choice
Web	SvelteKit 2 · Svelte 5 runes · Tailwind CSS 4 · shadcn-svelte · layerchart
Runtime	Bun (dev and production, via svelte-adapter-bun)
Data	PostgreSQL · Drizzle ORM · versioned migrations applied on boot
Mobile	Kotlin Multiplatform shared core · Jetpack Compose · SwiftUI · Ktor · SQLDelight · Koin
Offline	Dexie (web) · SQLDelight (Android) · SwiftData (iOS) · idempotent sync queue
AI	Model Context Protocol SDK · Apple Foundation Models (on-device)
Quality	Vitest · Playwright · Testcontainers · CodeQL · Semgrep · Trivy · Gitleaks

Repository layout

src/               SvelteKit app — routes, API, server logic, Drizzle schema
  lib/server/      auth, validation, MCP server, sync, security
mobile/
  shared/          Kotlin Multiplatform core (models, API client, repositories, DI)
  androidApp/      Jetpack Compose app
  wearApp/         Wear OS app
  iosApp/          SwiftUI app + widgets + Apple Watch app
drizzle/           SQL migrations and snapshots
crawler/           base food catalog importer
tests/             unit, integration (Testcontainers) and Playwright e2e suites
analytics-parity/  golden vectors keeping the TS and Kotlin analytics in step
scripts/           codegen, security scans, release helpers
store/             App Store and Play Console metadata
docs/              generated OpenAPI spec and the MCP guide

Development

Requires Bun, PostgreSQL, and OAuth credentials for at least one sign-in provider (see .env.example).

bun install
cp .env.example .env       # fill in DATABASE_URL, SESSION_SECRET, an OIDC provider
bun run dev                # migrations run automatically on start

bun run check              # svelte-check + prettier
bun test                   # unit tests
bun run test:integration-db # DB integration tests (Testcontainers, needs Docker)
bun run test:mobile        # Playwright e2e
bun run api:generate       # regenerate OpenAPI spec + TS/Kotlin clients
bun run analytics:check    # verify the TS/Kotlin analytics parity vectors
bun run constants:check    # verify the generated shared constants
bun run security           # Semgrep + bun audit + Trivy

Android:

cd mobile && ./gradlew androidApp:assembleDebug

iOS requires macOS with Xcode; the shared KMP module builds as a static framework.

Support

Bissbilanz is free, with no ads and nothing to unlock. If it is useful to you, you can help cover hosting on the support page or via GitHub Sponsors — entirely voluntary, and it changes nothing in the app. Bugs and requests go to issues; security reports follow SECURITY.md.

License

PolyForm Noncommercial 1.0.0 — free to use, modify and share for any noncommercial purpose. Commercial use requires a separate license.

---

_{Built by Orell Bühler · nutrition data from Open Food Facts}