An AI-powered cryptocurrency trading bot with reinforcement learning (RL) capabilities, integrated with OpenClaw agent memory and context.
Latest Performance: +27.69% gain in 9 hours (3 SHORT positions closed on 2026-02-28)
main_trading_loop.py- Main bot entry pointpaper_trading.py- Paper trading simulator (no real money)trading_bot.py- Live trading implementation- Training scripts for RL model fine-tuning
- Episode monitoring and automated learning
memory/- Agent memory logs (2026-02-21 to 2026-02-28)AGENTS.md- Agent workspace guidelinesSOUL.md- Agent personalityUSER.md- User contextTOOLS.md- Tool configurationIDENTITY.md- Agent identityHEARTBEAT.md- Proactive checks
# Install dependencies
python3 -m venv venv
source venv/bin/activate
pip install -r requirements_trading.txt
# Configure API keys
cp .env.example .env
# Edit .env with your Binance API keys
# Run paper trading (safe mode)
python3 main_trading_loop.pyModel: DeepSeek-R1-Distill-Llama-8B
Risk Management:
- Position size: 30-40% per trade
- Leverage: 10x
- Stop Loss: 5% | Take Profit: 10%
- Max daily trades: 3
Supported Pairs: BTC/USDT, ETH/USDT, SOL/USDT
2026-02-27 to 2026-02-28:
- 3 SHORT positions closed profitably
- Total P&L: +$27.69 USDT (+27.69%)
- Starting balance: $100 → Final: $127.69
Details in memory/2026-02-28.md
USE AT YOUR OWN RISK. Cryptocurrency trading is highly volatile. Start with paper trading. No warranties provided.
TRADING_README.md- Trading bot overviewAGENT_LIGHTNING_PLAN.md- RL training planEPISODE_MONITORING_GUIDE.md- Episode collectionPAPER_TRADING_TEST_SUMMARY.md- Test results
Includes full agent memory and context for seamless deployment on any system.
Created: 2026-02-28 | Status: Experimental - Paper Trading Validated ✅