Sharingan

AI-powered portfolio management system for Indian equities. Built as Claude Code skills — scans markets, analyses stocks, monitors positions, and executes trades via Groww API.

The eye that reads the market.

Architecture

/market-scan          →  Deep research, finds opportunities  →  watchlist.json
/portfolio-sweep      →  Monitors positions, proposes trades  →  portfolio.json
stock_analyser.py     →  Hybrid yfinance + Groww analysis engine (TA-Lib)
trading_client.py     →  Groww API wrapper for execution

Data Sources

Source	What	Speed
yfinance	Fundamentals, historical OHLCV, screening	Free, 15-min delayed
Groww API	Real-time quotes, market depth, OI, options, order execution	Paid, real-time
TA-Lib	RSI, MACD, Bollinger, ADX, Stochastic, S/R, Fibonacci, candlestick patterns	Local computation

Setup

1. Python Environment

conda create -n pygeneral python=3.13
conda activate pygeneral
pip install yfinance pandas numpy growwapi pyotp
# TA-Lib requires system install first: brew install ta-lib
pip install TA-Lib

2. Groww API Credentials

mkdir -p ~/.claude/trading
cat > ~/.claude/trading/.env << 'EOF'
GROWW_API_KEY=your_totp_token_jwt
GROWW_API_SECRET=your_totp_base32_secret
EOF

Get keys from: https://groww.in/trade-api (use TOTP flow for no-expiry auth)

3. Install Skills

cp skills/*.md ~/.claude/commands/

4. State Files

mkdir -p ~/.claude/trading

echo '{"holdings": {}, "last_updated": ""}' > ~/.claude/trading/portfolio.json
echo '{"last_scan": "", "macro_theme": "", "stocks": {}}' > ~/.claude/trading/watchlist.json
echo '{"trades": []}' > ~/.claude/trading/trade_log.json
cat > ~/.claude/trading/risk_rules.json << 'EOF'
{
  "max_position_pct": 5,
  "max_stock_allocation_pct": 10,
  "max_open_positions": 10,
  "default_stop_loss_pct": 7,
  "trail_sl_to_breakeven_after_pct": 5,
  "daily_loss_limit_pct": 3,
  "no_averaging_down": true,
  "max_sector_allocation_pct": 30,
  "min_market_cap_cr": 500
}
EOF

Usage

# Daily research scan
claude "/market-scan"

# Frequent position check
claude "/portfolio-sweep"

# Direct analysis
python stock_analyser.py ONGC TCS RELIANCE --mode full

# Trading client
python trading_client.py holdings
python trading_client.py quote ONGC
python trading_client.py buy ONGC 25 --price 265
python trading_client.py option-chain NIFTY 2026-03-17

Workflow

1. /market-scan (daily) — WebSearch for macro themes, find second-order beneficiaries, run technicals, update watchlist
2. /portfolio-sweep (frequent) — Check holdings against SL/targets, scan watchlist for entry signals, propose trades
3. User approves — Never executes without explicit approval
4. State persists — All decisions logged in trade_log.json for learning

Disclaimer

This is a personal trading tool, not financial advice. Use at your own risk.