Lockean Lite is an autonomous SPY options paper-trading agent built around separation of powers. The AI may decide what it wants to trade, but it cannot grant itself permission to risk capital or access Alpaca directly. The agent continuously evaluates real Alpaca SPY market data, official Cboe VIX data, and live SPY option candidates while the market is open. It can autonomously choose TRADE or NO_TRADE. When it proposes a trade, trusted pricing and maximum loss are reconstructed independently before a separate Lockean Authority evaluates evidence, proposal integrity, account eligibility, buying power, and risk policy. An approved trade receives a cryptographically authenticated, short-lived Authorization Receipt bound to the exact proposal fingerprint. A separate Execution Gateway must independently verify that receipt before it can submit an Alpaca paper MLEG order. The AI cannot authorize itself, the Authority cannot reach the broker, and the Gateway cannot self-authorize. Lockean now runs continuously against the linked Alpaca paper account, supports up to five committed defined-risk spread positions, counts pending orders toward the portfolio cap, applies a daily-loss halt, and fails closed on transient errors. A live Streamlit Control Room reads the same Alpaca account every 15 seconds, displaying equity, total P&L, daily P&L, unrealized P&L, buying power, open positions, recent broker orders, and market status. In verified paper execution, Alpaca independently confirmed a SPY 782/785 multi-leg order FILLED at a $0.89 limit price. Controlled safety proof also demonstrated rejection when trusted maximum loss reached $164 against a $150 policy limit. The current engineering baseline is 213 automated tests passing with zero regressions. Lockean Lite proves that an AI trading agent can operate autonomously without ever becoming its own execution authority.
Category tags: