Options-M: Autonomus AI Trading Agent

options-m is an autonomous options-trading system built around five specialised agents that run concurrently in a single asyncio process, coordinating through a shared Postgres database. One agent scans a fixed universe of ten liquid stocks every minute and scores them for trade opportunity using deterministic technical signals (RSI extremity, realised volatility, and the IV/RV edge). Another continuously tracks open positions and computes live profit and loss. A strategist agent makes a single bounded AI call per cycle to assess market regime, then applies a deterministic rule table to choose an options strategy (such as a credit spread). An execution agent turns that strategy into a concrete order, runs pre-trade risk checks, and submits it to a live brokerage account, while a reflection agent reviews closed trades afterward and writes lessons back into the system to improve future decisions. Only two of the five agents touch AI at all, and even then for a narrow, bounded task, everything else, including risk limits, exit conditions, and contract selection, is pure rule-based code that is fully deterministic and reproducible. The defining design goal is auditability: every decision the system makes, including the trades it chooses not to take, is stored together with its complete reasoning chain, so any outcome can be traced back step by step. A companion Next.js dashboard and a local CLI both let a human inspect the system's live state, trace a decision, or dry-run a trade plan without ever risking an unintended order.