Duolingo for Finance — built for students and young adults who want to build real financial habits, not just read about them.
Most finance apps fail young adults in two ways.
They either lecture (dense content nobody reads) or track (dashboards nobody checks). Neither changes behavior.
People already know they should save. Telling them again doesn't work. The real problem is building something that makes better financial habits feel automatic — not effortful.
Four things working together in one app:
1. Automatic expense tracking — SMS and notification parsing captures every UPI/bank transaction in the background. No manual entry. No friction.
2. Nova (AI financial companion) — not a chatbot, a coach. Nova knows your income, spending patterns, and goals. It gives specific, actionable advice based on your actual behavior, not generic tips.
3. Zero-risk stock simulator — practice investing with ₹15,00,000 in virtual money using live Yahoo Finance data. Make mistakes without losing real money.
4. Gamified financial education — XP, streaks, and interactive quizzes. The same behavioral psychology that makes Duolingo addictive, applied to finance.
Most AI assistants in finance apps are glorified FAQ bots.
Nova is prompt-engineered to behave like a personal CFO who has read your bank statements. It combines your Financial Score (a 0–100 rating based on savings ratio, spending velocity, and goal adherence) with your recent transaction data to give advice that is specific to you.
Example Nova insight:
"You spent 30% of your income on dining this week. Cook at home for 3 days and your score goes up 5 points — keeping your savings goal on track."
Not generic. Not a chart. A specific next action.
Muscle Money is a full-stack monorepo built with Turborepo. muscle-money/
├── apps/
│ ├── mobile/ # Flutter app (iOS + Android)
│ └── api/ # NestJS core API
├── packages/
│ ├── ai-service/ # FastAPI — Nova AI orchestration
│ └── shared/ # Shared types and utilities
├── infrastructure/ # Docker + deployment configs
└── redis/ # Session and cache layer Mobile: Flutter + Dart + Riverpod + GoRouter Clean Architecture with strict separation of Presentation / Domain / Data layers.
Backend: NestJS (TypeScript) for core API, FastAPI (Python) for AI orchestration and ML features. Turborepo monorepo — shared types across services.
AI Layer: OpenRouter + Gemini 1.5 Flash. Nova is not a generic LLM call — it has a specific system prompt, financial context injection per user, and guardrails that keep it focused on finance only.
Automated Tracking: Android Notification Listener
- SMS reading (with explicit user permission). Regex parsing extracts Amount, Merchant, and Transaction Type from bank/UPI messages instantly.
Market Data: Yahoo Finance API for live stock prices and historical charting in the simulator.
Why offline-first local storage? Financial data needs to be available instantly — waiting for a network call to show your balance breaks the experience. Local-first with background sync keeps the app fast regardless of connectivity.
Why a monorepo? Shared TypeScript types between the Flutter app and NestJS backend eliminate a whole class of bugs where the frontend and backend disagree on data shapes.
Why FastAPI separate from NestJS? AI inference is Python-native. Keeping it in a separate FastAPI service means the Node.js API never blocks on AI calls — and the AI layer can be scaled independently.
| Feature | Description |
|---|---|
| Automatic Expense Tracking | SMS + notification parsing, zero manual input |
| Nova AI Coach | Context-aware financial guidance based on your actual data |
| Financial Score | 0–100 dynamic rating: savings ratio + spending velocity + goal adherence |
| Stock Simulator | Paper trading with live Yahoo Finance data + ₹15L virtual balance |
| Time Machine | Projects your portfolio 10–15 years forward using compound interest |
| Learning Hub | XP + streaks + quizzes — Duolingo-style finance education |
| Onboarding Personalization | Income, goals, debt, risk appetite → personalized from day one |
Flutter SDK 3.x
Node.js 18+
Python 3.10+
Docker (for Redis)git clone https://github.com/codeharsh27/Muscle-money.git
cd Muscle-money
npm install # installs all monorepo packagescp .env.example .env
# Add your keys:
# OPENROUTER_API_KEY
# YAHOO_FINANCE_API_KEY
# SUPABASE_URL + SUPABASE_KEY# Start Redis
docker-compose up -d redis
# Start NestJS API
cd apps/api && npm run dev
# Start FastAPI AI service
cd packages/ai-service && uvicorn main:app --reload
# Start Flutter app
cd apps/mobile && flutter runHarsh Mule — Product Engineer
I built Muscle Money because I watched friends lose track of money every month not because they didn't care, but because existing tools made tracking feel like work.
Nova exists because financial advice shouldn't be generic. It should know you.