ai-hedge-fund

name	ai-hedge-fund
description	An AI-powered hedge fund team that simulates legendary investors (Buffett, Munger, Graham, etc.) to analyze stocks and provide investment recommendations using multi-agent consensus.
homepage	https://github.com/virattt/ai-hedge-fund
metadata	{"openclaw":{"emoji":"📈","requires":{"bins":["python3","pip3"]},"install":[{"id":"yfinance","kind":"pip","package":"yfinance","bins":[],"label":"Install yfinance (Yahoo Finance)"},{"id":"pandas","kind":"pip","package":"pandas","bins":[],"label":"Install pandas"},{"id":"numpy","kind":"pip","package":"numpy","bins":[],"label":"Install numpy"}]}}

AI Hedge Fund Skill

An AI-powered hedge fund team that simulates legendary investors to analyze stocks and provide investment recommendations.

Overview

This skill creates a team of AI agents, each embodying the investment philosophy of famous investors:

Classic Investment Agents (5)

• Warren Buffett - Value investing, wonderful companies at fair prices
• Ben Graham - Margin of safety, hidden gems
• Technical Analyst - Chart patterns and indicators
• Risk Manager - Risk metrics and position sizing
• Cathie Wood - Innovation and disruption

Enhanced Analysis Agents (4) - NEW in v2.1

• Earnings Analyst - EPS surprises, beat rates, earnings quality
• Wall Street Consensus - Analyst ratings, price targets, upside
• Macro Strategist - VIX, market regime, SPY/QQQ trends
• Dividend Investor - Yield, payout safety, dividend growth

Total: 9 agents analyzing each stock for comprehensive coverage

Quick Start

# Unified CLI - All features in one command
ai-hedge-fund analyze AAPL
ai-hedge-fund portfolio AAPL,MSFT,GOOGL --risk moderate
ai-hedge-fund backtest AAPL,MSFT --start 2023-01-01 --end 2024-01-01

# Or use individual commands
./ai-hedge-fund AAPL                    # Basic analysis
./ai-hedge-fund AAPL --detailed         # Detailed agent reasoning
./ai-hedge-fund AAPL --hot              # Include trending stocks
./ai-hedge-fund AAPL --rumor            # Include rumor scan
./ai-hedge-fund AAPL --hot --rumor      # Full intelligence

./portfolio-build AAPL,MSFT,GOOGL       # Portfolio construction
./backtester AAPL,MSFT --start 2023-01-01 --end 2024-01-01  # Backtest

# Market Intelligence
./scanner hot                           # Trending stocks & crypto
./scanner rumor                         # Market-wide rumor scan
./scanner rumor -t NVDA                 # Scan specific ticker
./scanner scan                          # Hot + Rumor combined

Data Sources

Free Tier (No API Key Required)

• Yahoo Finance - Real-time prices, basic financials (via yfinance)
• AAPL, GOOGL, MSFT, NVDA, TSLA have extended free data

Optional API Keys (for enhanced data)

# Add to ~/.openclaw/skills/ai-hedge-fund/.env
FINANCIAL_DATASETS_API_KEY=your_key  # Detailed financial metrics
ALPHA_VANTAGE_API_KEY=your_key       # Alternative data source

Architecture

User Request
    │
    ▼
┌─────────────────┐
│  Data Fetcher   │ ← Yahoo Finance / API
└────────┬────────┘
         │
    ┌────┴────┬────────┬────────┬────────┐
    ▼         ▼        ▼        ▼        ▼
┌───────┐ ┌───────┐ ┌───────┐ ┌───────┐ ┌───────┐
│Buffett│ │ Munger│ │Graham │ │Burry  │ │Cathie │ ← Parallel sub-agents
└───┬───┘ └───┬───┘ └───┬───┘ └───┬───┘ └───┬───┘
    │         │         │         │         │
    └─────────┴────┬────┴─────────┴─────────┘
                   │
          ┌────────▼────────┐
          │  Risk Manager   │
          └────────┬────────┘
                   │
          ┌────────▼────────┐
          │Portfolio Manager│ ← Final recommendation
          └─────────────────┘

Agent Details

Warren Buffett Agent

Philosophy: "It's far better to buy a wonderful company at a fair price than a fair company at a wonderful price."

Analysis Criteria:

• Return on Equity (ROE) > 15%
• Debt to Equity < 0.5
• Operating Margin > 15%
• Consistent earnings growth
• Durable competitive moat
• Margin of safety calculation

Signal: bullish/bearish/neutral with confidence (0-100)

Charlie Munger Agent

Philosophy: "The big money is not in the buying and selling, but in the waiting."

Analysis Criteria:

• Mental model checklist
• Rational capital allocation
• Shareholder-friendly management
• Circle of competence
• Long-term thinking

Ben Graham Agent

Philosophy: "In the short run, the market is a voting machine but in the long run, it is a weighing machine."

Analysis Criteria:

• Margin of safety (price < intrinsic value * 0.66)
• P/E ratio < 15
• P/B ratio < 1.5
• Current ratio > 2
• No earnings deficit in past 10 years

Technical Analyst

Analysis:

• Moving averages (20, 50, 200 day)
• RSI (Relative Strength Index)
• MACD
• Support/resistance levels
• Volume analysis

Risk Manager

Metrics:

• Volatility (standard deviation)
• Beta vs S&P 500
• Maximum drawdown
• Sharpe ratio
• Position size recommendations

Enhanced Analysis (NEW in v2.1)

Based on features learned from stock-analysis skill, we've added 4 new agents:

Earnings Analyst

Focus: Earnings surprises and quality

• EPS surprise analysis (actual vs expected)
• Historical beat rate (last 4 quarters)
• Earnings growth trends
• Example: "Beat by +5.2%, 3/4 quarters exceeded estimates"

Wall Street Consensus

Focus: Professional analyst opinions

• Consensus rating (strong_buy/buy/hold/sell/strong_sell)
• Number of covering analysts
• Price target vs current price
• Upside/downside potential
• Example: "19 analysts, consensus HOLD, +1.2% upside to target"

Macro Strategist

Focus: Market environment context

• VIX level (fear index)
• Market regime (bull/bear/choppy)
• SPY/QQQ 10-day trends
• Risk-off/risk-on indicators
• Example: "VIX 19.6 (elevated), choppy market, SPY flat"

Dividend Investor

Focus: Income and dividend safety

• Dividend yield analysis
• Payout ratio safety assessment
• Dividend growth history

Market Intelligence (NEW in v2.2)

🔥 Hot Scanner

Find trending stocks and crypto before they hit mainstream

Data Sources:

• CoinGecko Trending (top trending crypto)
• Yahoo Finance Movers (gainers/losers/high volume)

Usage:

# Standalone hot scanner
./scanner hot

# Include in stock analysis
./ai-hedge-fund TSLA --hot

Output Example:

🔥 HOT SCANNER - Trending Stocks & Crypto
Found 6 trending assets

📈 STOCKS:
  🟢 UBER    +3.18%
  🟢 NVDA    +2.45%

🪙 CRYPTO:
  🚀 ESP-USD    +46.91%
  📉 OP-USD     -14.47%

🔮 Rumor Scanner

Detect early signals, M&A rumors, and insider activity

Detection Types:

• 🤝 M&A: Merger, acquisition, takeover rumors
• 👤 Insider: Insider buying/selling activity
• ⬆️ Upgrade: Analyst upgrades, price target raises
• ⬇️ Downgrade: Analyst downgrades, target cuts
• 📊 Earnings: Earnings surprises, guidance changes
• 🤝 Partnership: New deals, collaborations

Confidence Levels:

• 🔴 High: Reputable source (Reuters, Bloomberg, CNBC)
• 🟡 Medium: Confirmed keywords ("announces", "SEC filing")
• ⚪ Low: Early signals, unconfirmed

Usage:

# Scan specific ticker
./scanner rumor -t TSLA

# Scan market-wide
./scanner rumor

# Include in analysis
./ai-hedge-fund NVDA --rumor

Output Example:

🔮 RUMOR SCANNER - Early Signals
Detected rumors for 3 tickers

📰 NVDA:
  🟡 ⬆️ [UPGRADE] Goldman raises target to $850...
     Source: MarketWatch | Confidence: medium

📰 TSLA:
  ⚪ 🤝 [PARTNERSHIP] Rumored deal with...
     Source: Twitter | Confidence: low

• Income rating (excellent/good/moderate/poor)
• Example: "2.8% yield, safe payout at 45%, dividend aristocrat"

Usage Examples

Basic Analysis

# Single stock
result = ai_hedge_fund.analyze("AAPL")
# Returns: {"signal": "bullish", "confidence": 75, "reasoning": "..."}

# Multiple stocks
results = ai_hedge_fund.analyze(["AAPL", "MSFT", "GOOGL"])

Detailed Report

{
  "ticker": "AAPL",
  "analysis_date": "2025-02-18",
  "agents": {
    "warren_buffett": {
      "signal": "bullish",
      "confidence": 85,
      "reasoning": "Strong ROE of 160%, excellent brand moat, consistent dividend growth"
    },
    "charlie_munger": {
      "signal": "bullish",
      "confidence": 80,
      "reasoning": "Rational capital allocation via buybacks, strong pricing power"
    },
    "ben_graham": {
      "signal": "neutral",
      "confidence": 50,
      "reasoning": "High P/E of 28 exceeds margin of safety threshold"
    },
    "technical": {
      "signal": "bullish",
      "confidence": 70,
      "reasoning": "Price above 200-day MA, RSI at 55 indicates healthy momentum"
    },
    "risk_manager": {
      "signal": "neutral",
      "confidence": 60,
      "reasoning": "Beta 1.2 indicates market correlation, moderate volatility"
    }
  },
  "consensus": {
    "signal": "bullish",
    "confidence": 73,
    "agreement": "4/5 agents bullish",
    "key_risks": ["High valuation", "Market correlation"],
    "recommendation": "Consider position size of 5-8% max"
  }
}

Configuration

Environment Variables

# ~/.openclaw/skills/ai-hedge-fund/.env

# LLM Configuration (uses OpenClaw default if not set)
OPENAI_API_KEY=sk-...  # Optional, falls back to kimi-k2.5

# Data Sources
FINANCIAL_DATASETS_API_KEY=...  # Optional, enhances data quality
ALPHA_VANTAGE_API_KEY=...       # Optional, alternative source

# Analysis Settings
MAX_AGENTS=12           # Number of agents to run
PARALLEL_MODE=true      # Run agents in parallel
CACHE_DURATION=3600     # Cache data for 1 hour

Custom Agents

Add your own investment style by creating a new agent file:

# ~/.openclaw/skills/ai-hedge-fund/agents/custom_agent.py

from typing import Literal

class CustomAgentSignal(BaseModel):
    signal: Literal["bullish", "bearish", "neutral"]
    confidence: int = Field(description="Confidence 0-100")
    reasoning: str = Field(description="Reasoning for the decision")

def custom_agent(state: AgentState, ticker: str) -> CustomAgentSignal:
    """Your custom analysis logic"""
    # Fetch data
    data = fetch_financial_data(ticker)
    
    # Analyze
    score = analyze_custom_metrics(data)
    
    # Return signal
    return CustomAgentSignal(
        signal="bullish" if score > 70 else "neutral",
        confidence=score,
        reasoning="Your reasoning here"
    )

Limitations & Disclaimers

⚠️ IMPORTANT: This tool is for educational and research purposes only.

• Not investment advice: These are AI simulations, not professional financial advice
• No guarantee: Past performance analyzed by AI does not predict future results
• Data limitations: Free data sources may have delays or inaccuracies
• Risk: Always consult a qualified financial advisor before making investment decisions

Technical Details

Multi-Agent Coordination

Uses OpenClaw's sessions_spawn to run agents in parallel:

// Parallel agent execution
const agentResults = await Promise.all(
  agents.map(agent => 
    sessions_spawn({
      task: `Analyze ${ticker} as ${agent.name}`,
      agentId: 'investment-analyst',
      timeoutSeconds: 60
    })
  )
);

Data Caching

• Financial data cached for 1 hour to reduce API calls
• Agent results cached for same ticker within 30 minutes

Error Handling

• Graceful fallback if data source unavailable
• Individual agent failures don't block other agents
• Missing data reported in reasoning

Troubleshooting

"No data for ticker"

• Check ticker symbol is correct (e.g., "BRK-B" not "BRK.B")
• Try popular tickers first (AAPL, MSFT, GOOGL)
• Some tickers may not be available in free tier

"API rate limit"

• Wait a few minutes and retry
• Results are cached, so retry is fast
• Consider adding API key for higher limits

Slow analysis

• First run fetches and caches data
• Subsequent runs use cached data (much faster)
• Use --quick flag for essential agents only

Feature Modules

1. Portfolio Construction (portfolio_constructor.py)

Modern Portfolio Theory (MPT) optimization:

• Mean-variance optimization
• Risk parity weighting
• Sector diversification analysis
• Sharpe ratio maximization
• Three risk profiles: conservative, moderate, aggressive

ai-hedge-fund portfolio AAPL,MSFT,GOOGL,JPM,JNJ --risk moderate

2. Backtesting (backtester.py)

Backtest strategies on historical data:

• Multiple strategies: ai_consensus, equal_weight, momentum, value
• Rebalancing schedules: weekly, monthly, quarterly
• Performance metrics: Sharpe, max drawdown, alpha, beta
• Benchmark comparison (S&P 500)
• Trade history tracking

ai-hedge-fund backtest AAPL,MSFT,GOOGL --start 2023-01-01 --end 2024-01-01 --strategy ai_consensus

3. Rebalancing Monitor (rebalance_monitor.py)

Monitor portfolio drift and generate alerts:

• Weight drift detection
• Signal-based target weights
• Urgency classification (HIGH/MEDIUM/LOW)
• Health score calculation
• Rebalancing schedule generation

ai-hedge-fund rebalance AAPL:0.3,MSFT:0.2,GOOGL:0.5 --last-rebalanced 2024-01-01

4. Tax Optimization (tax_optimizer.py)

Tax-loss harvesting and optimization:

• Unrealized gain/loss tracking
• Tax-loss harvesting opportunities
• Wash sale rule detection
• Replacement security suggestions
• Year-end tax strategy

ai-hedge-fund tax --lots '[{"ticker":"AAPL","shares":100,"purchase_date":"2024-01-01","purchase_price":150}]' --year-end

5. ESG Screening (esg_screener.py)

Environmental, Social, Governance screening:

• ESG score calculation (0-10 scale)
• Controversy detection
• Sector comparison
• Exclusion criteria checking
• Portfolio ESG scoring

ai-hedge-fund esg AAPL,MSFT,XOM,TSLA --portfolio --minimum-score 6.0

File Structure

ai-hedge-fund/
├── SKILL.md                      # This documentation
├── QUICKSTART.md                 # Quick start guide
├── ADVANCED.md                   # Advanced architecture
├── ai_hedge_fund.py             # Basic rule-based analysis
├── ai_hedge_fund_advanced.py    # AI-powered sub-agent analysis
├── portfolio_constructor.py     # Portfolio optimization
├── backtester.py                # Strategy backtesting
├── rebalance_monitor.py         # Rebalancing alerts
├── tax_optimizer.py             # Tax-loss harvesting
├── esg_screener.py              # ESG screening
├── ai-hedge-fund-cli            # Unified CLI
├── ai-hedge-fund                # Basic CLI wrapper
├── ai-hedge-fund-advanced       # Advanced CLI wrapper
├── portfolio-build              # Portfolio CLI wrapper
└── .env                         # API keys

Related Resources

• Original Project
• Value Investing
• Technical Analysis
• Modern Portfolio Theory
• ESG Investing

---

Version: 2.0.0
Author: OpenClaw Community
License: MIT