SentinelTrade

SentinelTrade is a modular, risk-gated trading agent built entirely on Alpaca's paper trading environment for the Alpaca AI Trading Agents Hackathon. The system is structured as a pipeline of independent components: 1. Market Scanner — pulls live daily bar data for a watchlist of symbols (SPY, QQQ, AAPL, MSFT, NVDA) via Alpaca's Market Data API. 2. Strategy Engine — calculates 5-day and 20-day moving averages, RSI, and a composite momentum score from the raw price data, then converts these into a transparent BUY/HOLD/SELL signal per symbol. 3. Risk Manager — acts as a hard, independent safety boundary that the strategy cannot override. It enforces a maximum position size (10% of account equity per symbol), a maximum total portfolio exposure (60%), a daily loss circuit breaker (2%), and a minimum trade size, rejecting or resizing any trade that breaks these rules. 4. Position-Aware Execution — before submitting any trade, the agent checks existing paper account holdings via Alpaca's Trading API, so it never re-buys a position it already holds, and calculates exposure against real account state rather than assuming a fresh start each run. 5. Dry-Run / Live Modes — every run defaults to a dry-run mode that prints exactly what would be submitted without sending any orders, with a `--live` flag required to actually route orders to Alpaca's paper trading endpoint. The core design principle is separation of concerns: the strategy proposes, but only the risk manager can approve. This makes the system's safety guarantees auditable independent of whatever logic drives the trading signal itself — a foundation we intend to build an LLM-based reasoning layer on top of next, where an AI model reviews each proposed trade's context and can add a layer of judgment before the risk manager has final say. All trading in this build was tested exclusively in Alpaca's paper environment with simulated funds — no real capital was placed at risk at any point.