Research backed options strategy

Most retail options tooling stops at the idea. It surfaces a signal or a headline and leaves the hardest parts — structure selection, sizing, and risk — to the trader. This project closes that loop with an agent that goes from research to a managed position without guessing. The pipeline has three layers. A Python quant layer ingests market and options data from Alpaca, builds the features behind each candidate setup, and scores them so that only setups clearing a defined statistical bar move forward. Claude then acts as the research analyst: it reasons over the context around the signal, checks the quantitative view against what is actually happening in the name, and writes an explicit thesis along with the conditions that would invalidate it. A signal the model cannot justify in plain language does not trade. Finally, an execution layer translates the view into a defined-risk options structure — spreads rather than naked exposure — sizes it against account risk limits, and routes it through the Alpaca API. Risk management is part of the strategy rather than a wrapper around it. Maximum loss is bounded at entry by construction, position size is a function of account equity and spread width, and every open position carries the exit conditions its thesis was written against, so the agent closes on invalidation instead of on hope. The signals were developed and validated in a Python backtesting layer before going live, and the agent is now trading them in live paper trading on Alpaca. Each trade is logged with both its signal inputs and its written rationale, which makes the strategy reviewable rather than a black box.