mahmoud20138 autohedge-swarm
USE FOR:
- "autonomous hedge fund agent"
- "swarm agents for trading"
- "director + quant + risk manager pipeline"
- "Solana autonomous trading"
- "enterprise AI trading system"
- "risk-first automated trading" tags: [swarm, agents, hedge-fund, autonomous, Solana, risk-and-portfolio, enterprise, multi-agent, quant] kind: framework category: quant-ml-trading
What Is AutoHedge?
Enterprise-grade autonomous agent hedge fund using swarm intelligence. Specialized agents sequentially handle strategy, analysis, risk, and execution.
- Repo: https://github.com/The-Swarm-Corporation/AutoHedge
- Install:
pip install -U autohedge - Current: Solana trading (Coinbase planned)
- Philosophy: Risk-first — position sizing happens before execution
Swarm Agent Pipeline
Director Agent
↓ Strategy generation + market context
Quant Agent
↓ Quantitative analysis + signal generation
Risk Manager Agent
↓ Position sizing + risk assessment + approval
Execution Agent
↓ Order construction + submission
Trade Output (JSON)
Installation
pip install -U autohedge
Environment variables:
JUPITER_API_KEY="..." # Solana DEX aggregator
OPENAI_API_KEY="sk-..." # or ANTHROPIC_API_KEY
WALLET_PRIVATE_KEY="..." # Solana wallet
Usage
autohedge
Or programmatically:
from autohedge import AutoHedge
fund = AutoHedge(
llm_provider="anthropic", # Director/Quant use Claude
risk_threshold=0.02, # Max 2% portfolio risk per trade
chain="solana",
)
result = fund.analyze_and_trade("SOL/USDC")
print(result) # JSON: analysis + decision + risk metrics
Agent Responsibilities
| Agent | Role |
|---|---|
| Director | Market context, strategy selection |
| Quant | Price analysis, signals, technicals |
| Risk Manager | Position sizing, max drawdown limits |
| Execution | Order construction, submission |
Output Format (JSON)
{
"ticker": "SOL/USDC",
"director_analysis": "Bullish momentum...",
"quant_signals": {"rsi": 58, "macd": "bullish"},
"risk_assessment": {"position_size": 0.015, "stop_loss": 0.02},
"decision": "BUY",
"execution": {"order_type": "market", "size": 10.5}
}
Key Design Principles
- Risk-first: Never execute without risk approval
- Audit trail: Enterprise logging at every step
- Modular: Swap any agent or add custom stages
- Structured outputs: All agents return JSON for system integration