Determinism | Autonomous Trader

A fully Autonomous agent that works all the way till the end the market day (4PM ET), requiring you to just press a button. Our core principle is Determinism controls the money, Agents provide the intelligence. The architecture starts with the universe, narrowed by a deterministic scanner using rule-based filters (information fetched using Alpaca). A Featherless LLM research agent then generates research (fundamentals, news, catalysts, risks, sourced from Alpaca) on the scanned companies, which feeds a second Stock Ranker Agent that re-ranks and narrows this list to the top N stocks. Notice, how we focus on RANKING over SCORING. This is our advantage. For each selected company, a deterministic Black-Scholes-Merton Option Selector picks 5 option contracts based on quantitative criteria (delta, DTE, moneyness, liquidity, IV, provided by Alpaca). Across companies, these selections pool into a common option pool (e.g., 50 companies × 5 options = 250 contracts). An LLM option-ranking agent then evaluates this entire pool using both market data and the earlier company research to surface the best opportunities. From there, a deterministic layer builds the portfolio, runs risk evaluation (Greeks, VaR, concentration, equity data extracted from Alpaca). Our deterministic risk makes sure your equity stays stable and doesn’t fall hard, and finally executes (w/ Alpaca API) and confirms trades. The agent loops and keeps track of your running trades ensuring good exit, and adds new trades over the trading day automatically periodically. Here, LLMs handle research and ranking, deterministic logic handles selection and execution. SDK Reasoning: As part of the end points of the pipeline are based in determinism, we used alpaca-py SDK over MCP, and over CLI for just programming convenience.