Aegis Alpha

Aegis Alpha is an adversarial AI Investment Committee for autonomous options trading, built around Alpaca paper-trading infrastructure. Most AI trading agents focus on generating trade ideas. Aegis Alpha focuses on the decision boundary before execution: should this trade be allowed to risk capital at all? A proposed defined-risk options trade moves through market direction analysis, options strategy construction, deterministic risk checks, and a mandatory adversarial Investment Committee. The committee includes a Market Regime Agent, Options Quant Agent, Bull Advocate, Bear Advocate, and deterministic Risk Officer. The Bull Advocate builds the strongest evidence-backed case for the trade, while the Bear Advocate actively challenges the thesis and surfaces failure conditions. The Risk Officer applies deterministic risk rules and has absolute veto authority. AI can advise, explain, and debate, but it cannot override risk controls. Aegis Alpha supports APPROVE, REJECT, and ABSTAIN outcomes. Abstention is treated as a valid safety outcome when evidence is insufficient or disagreement remains unresolved. A high-confidence signal alone is not enough to release capital. Every committee decision is auditable through structured reasoning, replayable decision records, counterfactual explanations, and tamper-evident SHA-256 decision certificates. For the hackathon demonstration, Aegis Alpha runs in paper mode, uses defined-risk options workflows, exposes a read-only Judge Mode dashboard, and keeps execution dry-run/safe by default. The public demo includes deterministic scenarios showing strong approval, risk veto, disagreement/abstention, and neutral/no-trade outcomes. The project demonstrates a safer model for autonomous financial agents: adversarial reasoning, deterministic risk authority, abstention, auditability, and governed capital release.