🤝 Top Collaborators
🤓 Latest Submissions
TradeCouncil — Multi-Agent Options Alpha
Most trading-agent demos ask one model to analyze everything and place an order. That spends inference credits indiscriminately, hides disagreement, and lets generated text get too close to capital. TradeCouncil separates those responsibilities. Python first ranks a liquid watchlist using trend, moving averages, RSI, momentum, and volatility. Only the top one or two candidates reach the AI council, keeping the default cycle within six Featherless calls. A Bull Analyst builds the upside case, a Bear Analyst challenges it, and a CIO Judge returns a direction, confidence score, and rationale. Deterministic code then selects an eligible call or put and applies hard limits for DTE, delta, bid/ask spread, open interest, premium-at-risk, drawdown, and portfolio exposure. A model cannot override a Risk Sentinel veto. The full decision trail—quant ranking, adversarial debate, CIO verdict, contract, risk result, and execution receipt—is visible in one dashboard. The local competition workflow uses Alpaca market/trading APIs and the official Alpaca CLI against a dedicated paper account. The public judge walkthrough is intentionally credential-free, synthetic, and non-mutating, so anyone can inspect the experience without receiving broker authority. TradeCouncil's product path is a paper-first research and review workspace with saved risk policies, decision journals, broker reconciliation, and team analytics. It does not promise returns or passive income; options can expire worthless and involve substantial risk.
4 Sep 2026
