Options Alpha Agent

Options Alpha Agent runs three independently-decisioned autonomous options-trading agents on three separate Alpaca paper accounts, all bounded by one shared, hard-coded risk layer: a kill switch checked first always, a market-open and macro-event blackout calendar, defined-risk-only structures, a per-trade risk cap, an aggregate risk cap, a daily entry limit, and a daily-loss circuit breaker. None of these are prompt instructions — they're enforced in code before any order reaches Alpaca's API, so no model can reason its way around them. Company A trades a deterministic 2-of-4 technical-criteria gate with no AI in the decision loop. Company B is the fully autonomous entry: a LangChain agent backed by an LLM reasons over live market signals and proposes trades directly through tool access — but every proposal still passes an independent second model call that can only vote yes or no, with zero tool access of its own, before the same guardrails see it. Company C runs a volatility-edge strategy: it computes its own Black-Scholes implied volatility and delta from live quoted prices, since Alpaca's own options feed returns all-zero greeks, buys puts only when implied volatility is cheap relative to realized volatility, and delta-hedges with the underlying stock. Execution runs entirely through Alpaca's own CLI, on a scheduler checking every account unattended for the full trading week. Every gate check, model verdict, and order is written to an audit log, visible live on a Streamlit dashboard alongside real-time equity, P&L, and positions for all three accounts side by side.