.png&w=256&q=75)
🤓 Latest Submissions
Autonomous Multi-Agent Options Trading System
Autonomous Multi-Agent Options Trading System An institutional-grade quantitative options trading engine built for the Alpaca AI Trading Agents Hackathon. The system orchestrates 4 specialized autonomous sub-agents communicating over strongly typed Pydantic contracts, utilizing Featherless AI serverless LLMs and the Alpaca Model Context Protocol (MCP) Server with Level-3 OrderClass.MLEG execution on a $100,000 Paper Trading account. 1. Multi-Agent Architecture: - Market Intelligence Agent: Ingests OCC option chains and equity quotes (SPY, QQQ, AAPL, NVDA, TSLA, MSFT). Computes Black-Scholes Greeks and Brent IV Rank, paired with Featherless AI (Meta-Llama-3.1-8B) for sub-320ms catalyst sentiment scoring. - Strategy and Options Agent: Synthesizes defined-risk spread structures: Bull Call Spreads, Bear Put Spreads, Iron Condors (range-bound theta capture), and Long Straddles (volatility breakouts). - Risk Management Agent (The Guardrail): 100% deterministic Python rules engine (Zero LLM hallucinations). Executes 6 quantitative gates in 1.5 microseconds: 3% NAV position cap ($3,000) with auto-downsizing, -3% daily drawdown breaker (-$3,000), 20% margin ceiling ($20,000), liquidity filter (OI>=50, spread<15%), min risk/reward >= 1.20 (vetoing naked options), and portfolio |Delta| <= 150. - Execution Agent: Dispatches atomic Level-3 MLEG orders via Alpaca MCP tools and manages position lifecycles (+80% Take-Profit / -50% Stop-Loss). 2. Performance & Infrastructure: - Audited Metrics: 55.4% Win Rate, 1.73 Profit Factor, +1.82 Simulated Sharpe Ratio, and $1.69M capital preserved via deterministic risk vetoes. - Zero-Polling Streaming: Alpaca WebSocket feeds with 38 microsecond in-memory quote caching (3.3s 6-symbol market sweep). - Mission Control UI: Glassmorphic FastAPI & WebSocket dashboard with dynamic SVG payoff curves and live telemetry.
4 Sep 2026
