Multi-Agent Financial Market Analysis System

A sophisticated multi-agent AI system that combines news sentiment analysis, technical market indicators, and cross-verification logic to provide intelligent trading insights. Built for hackathon submission targeting Anthropic (Best Use of Claude) and Apify tracks.

TradrFin uses a 4-Agent Multi-Agent System to analyze financial markets:

┌─────────────────────────────────────────────────────────────┐
│                    Frontend Dashboard                       │
│                    (React + Tailwind CSS)                   │
└───────────────────────┬─────────────────────────────────────┘
                        │
        ┌──────────────────────────────┐
        │                              │
┌───────▼──────┐                 ┌─────▼──────────┐
│  Backend API │                 │  Apify Actor   │
│  (Node.js)   │                 │  (Sentiment)   │
└───────┬──────┘                 └─────┬──────────┘
        │                              │
        └───────┬───────────────┬──────┘
                │               │
    ┌───────────▼───────────────▼───────────────┐
    │         Multi-Agent System                │
    ├───────────────────────────────────────────┤
    │  Agent 1: News & Social Sentiment Agent   │
    │  Agent 2: Asset Movement Diagnostics      │
    │  Agent 3: Cross-Verification Agent        │
    │  Agent 4: User-facing Advisor Agent       │
    └───────────────────────────────────────────┘

Purpose: Understand market mood from news and social signals

Inputs:

• Financial news headlines (Google News RSS, Reuters RSS)
• Market-wide sentiment indices
• Social sentiment data

Outputs:

• Overall sentiment: Bullish/Bearish/Neutral
• Sector-level sentiment (tech, finance, energy)
• Key drivers (inflation, earnings, geopolitics)
• Confidence score (0-1)

Data Sources:

• Finnhub News Sentiment API
• Google News RSS feeds
• Reuters RSS feeds
• Apify Actor for data collection

Implementation:

• Location: agents/news_sentiment_agent/
• Uses Claude (Anthropic) for sentiment analysis
• Deployed as Apify Actor

Purpose: Analyze real market behavior through technical indicators

Inputs:

• Price candles (1d or 1h)
• Volume changes
• Volatility (ATR - Average True Range)
• Momentum indicators (RSI, MACD)
• Market indices (S&P500, NASDAQ, BTC)

Outputs:

• Market trend assessment (upward/downward/neutral)
• Momentum strength
• Volume analysis
• Asset-specific signals (AAPL, TSLA, GOOG, etc.)

Technical Indicators:

• Trend/Momentum: 20-day MA vs 50-day MA
• Volatility: ATR (Average True Range)
• Volume: Today's volume vs 5-day moving average
• RSI: 51-70 = bullish, 30-50 = neutral, <30 = oversold

Data Sources:

• yfinance (free tier)
• AlphaVantage (free tier)
• Finnhub (free tier)

Implementation:

• Location: agents/decision agent/
• Precomputes indicators via precompute_indicators.py
• Analyzes historical data for multiple symbols

Purpose: Validate sentiment against actual price action - the "magic" that adds intelligence

Logic Examples:

Scenario 1: Contradiction Detection

Sentiment = Bullish
Price Trend = Down
Volume = Low
→ Output: "Sentiment is positive but price action contradicts it. 
           Market may be in distribution or false hype."

Scenario 2: Contrarian Setup

Sentiment = Bearish
Price Trend = Up
Volatility = High
→ Output: "Market is climbing despite fear. 
           Likely short squeeze or contrarian setup."

Outputs:

• Final combined sentiment (Bullish/Neutral/Bearish)
• Confidence score (0-1)
• Plain English explanation
• 3 key reasons for the assessment

Implementation:

• Location: agents/decision agent/decision_agent.py
• Uses Claude LLM for intelligent cross-verification
• Rule-based fallback mode available

Purpose: Transform all internal signals into actionable insights for users

Outputs:

• Clean, readable summary
• Actionable trading insights
• Visual signals (Green/Bullish, Red/Bearish, Yellow/Neutral)
• Beginner-friendly explanations

Implementation:

• Location: backend/server.js (POST /api/chat)
• Frontend: frontend/src/pages/Advisor.jsx
• Provides conversational interface

Market Sentiment: Bearish (score: 0.32)
Asset Trend: Bullish (score: 0.65)
Final Verdict: Neutral – conflicting signals
Confidence: 0.41

Explanation:
Headlines show strong concerns about inflation and slowing earnings.
BUT the S&P500 is in a short-term uptrend with strong buying volume.
This is often seen during short-covering or temporary relief rallies.

Recommendation:
Wait for confirmation before making large trades.

• React 19.2.0 - UI framework
• Vite - Build tool with Rolldown
• Tailwind CSS - Styling
• React Router - Navigation
• Recharts - Data visualization

• Node.js/Express - FinTech Dashboard API (port 5000)
• Python/FastAPI - Trading Simulation API (port 8000)
• MySQL - Cloud database for sentiment and market data

• Claude (Anthropic) - Sentiment analysis and decision-making
• Apify - News collection and data processing
• yfinance - Market data fetching
• Pandas - Data processing

• Apify Actor - News sentiment collection
• Python Agents - Decision making and execution
• FastAPI - Agent orchestration

• Node.js 16+
• Python 3.8+
• MySQL database (cloud or local)
• Anthropic API key
• Apify account (optional, for actor deployment)

1. Clone the repository:

   git clone <repository-url>
   cd TradrFin

2. Setup Backend (Node.js):

   cd backend
   npm install

   Create .env file:

   DB_HOST=your-database-host.com
   DB_USER=your-username
   DB_PASSWORD=your-password
   DB_NAME=your-database-name
   DB_PORT=3306
   DB_SSL=true
   PORT=5000

3. Setup Backend (Python):

   cd backend
   pip install -r requirements.txt

4. Setup Frontend:

   cd frontend
   npm install

5. Setup Sentiment Agent (Apify):

   cd agents/news_sentiment_agent
   npm install

6. Precompute Market Indicators:

   cd agents/decision\ agent
   python precompute_indicators.py

Terminal 1 - Node.js Backend:

cd backend
npm start
# Server runs on http://localhost:5000

Terminal 2 - Python FastAPI Backend:

cd backend
uvicorn app:app --reload --port 8000
# Server runs on http://localhost:8000
# API docs at http://localhost:8000/docs

Terminal 3 - Frontend:

cd frontend
npm run dev
# App runs on http://localhost:5173

TradrFin/
├── agents/
│   ├── decision agent/          # Agent 2 & 3: Market analysis & cross-verification
│   │   ├── decision_agent.py    # Cross-verification logic with Claude
│   │   ├── execution_agent.py    # Trade execution
│   │   ├── trading_simulator.py  # Simulation engine
│   │   └── precompute_indicators.py  # Technical indicator computation
│   └── news_sentiment_agent/     # Agent 1: Sentiment collection (Apify Actor)
│       ├── src/main.js           # Apify Actor implementation
│       └── actor.json            # Actor configuration
├── backend/
│   ├── server.js                # Node.js API (Agent 4: Advisor)
│   ├── app.py                    # FastAPI (Trading simulation)
│   └── requirements.txt          # Python dependencies
├── frontend/
│   ├── src/
│   │   ├── pages/
│   │   │   ├── Advisor.jsx       # Agent 4 UI
│   │   │   ├── Sentiment.jsx     # Agent 1 visualization
│   │   │   └── Technicals.jsx    # Agent 2 visualization
│   │   └── components/           # Reusable components
│   └── package.json
└── README.md

• GET /api/health - Health check
• GET /api/sentiment - Get sentiment analysis data
• GET /api/market-data - Get market technical data
• GET /api/transactions - Get transaction history
• POST /api/chat - AI chat endpoint (Agent 4)

• GET / - API information
• POST /demo - Run trading simulation
• GET /portfolio - Get current portfolio state
• GET /status - Get simulation status
• GET /logs - List trading log files

See backend/README.md for detailed API documentation.

✅ Multi-Agent Architecture - 4 specialized agents working together
✅ Claude AI Integration - Advanced sentiment analysis and decision-making
✅ Apify Actor - Scalable news collection and processing
✅ Cross-Verification Logic - Validates sentiment against price action
✅ Real-time Market Analysis - Technical indicators and trend detection
✅ Interactive Dashboard - Beautiful React UI with real-time updates
✅ Trading Simulation - Backtest strategies with historical data
✅ Confidence Scoring - Transparent AI decision-making

1. Agent 1 collects financial news and analyzes sentiment using Claude
2. Agent 2 computes technical indicators (RSI, MACD, ATR, Volume)
3. Agent 3 cross-verifies sentiment vs. price action, generates final signal
4. Agent 4 presents insights in user-friendly format via chat interface

The system flags contradictions (e.g., bullish sentiment but bearish price action) and provides explanations, making it valuable for both beginners and experienced traders.

cd agents/news_sentiment_agent
apify run --input='{"query":"NVIDIA","maxNewsItems":10}'

cd agents/decision\ agent
python test_decision_agent.py

cd agents/decision\ agent
python test_backtester.py

The system expects these tables in your MySQL database:

• sentiment_agent - Sentiment analysis results
• market_data_agent - Technical market data
• asset_movement_agent - Transaction records

• / - Banking Dashboard
• /advisor - AI Financial Advisor (Agent 4)
• /sentiment - Market Sentiment Analysis (Agent 1)
• /technicals - Technical Analysis (Agent 2)
• /ai-invest - AI Investment Tools
• /history - Transaction History
• /transfers - Money Transfers

DB_HOST=your-database-host
DB_USER=your-username
DB_PASSWORD=your-password
DB_NAME=your-database-name
DB_PORT=3306
DB_SSL=true
PORT=5000

ANTHROPIC_API_KEY=your-anthropic-api-key

• ✅ Agent 1: News & Social Sentiment (Apify Actor)
• ✅ Agent 2: Asset Movement Diagnostics
• ✅ Agent 3: Cross-Verification Agent
• ✅ Agent 4: User-facing Advisor
• ✅ Frontend Dashboard
• ✅ Backend APIs
• ✅ Trading Simulation

• backend/README.md - Detailed backend API documentation
• frontend/README.md - Frontend setup and structure
• agents/news_sentiment_agent/README.md - Apify Actor documentation
• agents/decision agent/DESIGN_PLAN.md - Agent design details

This is a hackathon project. For questions or improvements, please open an issue or submit a pull request.

This project is part of a hackathon submission.

• Best Use of Claude: Leverages Claude for intelligent sentiment analysis and cross-verification
• Apify Integration: Custom Apify Actor for scalable news collection
• Multi-Agent Design: 4 specialized agents working in harmony
• Real-world Application: Practical trading insights with confidence scoring
• Judge-friendly Output: Clear explanations and actionable recommendations

Built with ❤️ for the hackathon