ConvictionSpread: Explainable Options Agent

My friend and I built ConvictionSpread because we wanted an options trading agent that could explain itself before taking a position. The agent studies completed market data and evaluates trend, momentum, realized volatility, relative volume, ATR and relative strength. It uses these signals to develop a bullish, bearish or pass thesis. When the evidence is strong enough, it selects a defined-risk bull call or bear put debit spread from Alpaca’s options market data. A structured thesis-and-critic layer explains the supporting evidence, conflicting signals and conditions that would invalidate the trade. However, the AI never has final authority over execution. Deterministic safety checks independently validate market hours, quote freshness, liquidity, options buying power, position limits, maximum loss and account reconciliation. ConvictionSpread uses Alpaca’s Trading API and paper environment for market data, option discovery, multi-leg orders and broker-confirmed performance reporting. We also created a restricted Alpaca MCP configuration for structured, read-only account access. The project includes conservative bid-and-ask execution assumptions, fees, slippage, walk-forward testing and comparisons with buy-and-hold, momentum and random-direction baselines. To prevent look-ahead bias, every feature uses only information available at that point in time. The complete entry-to-exit lifecycle is supported, including exact order previews, one-contract paper canaries, durable order tracking, broker reconciliation, controlled spread closing and safe recovery from network timeouts without duplicate submissions. The project currently has 102 automated tests and a deployed decision dashboard. It remains paper-trading only. Our goal is not to claim guaranteed profitability, but to demonstrate an autonomous, transparent and safety-focused way to research and execute directional options strategies.