Hermes Trader — Autonomous AI Trading Agent

Hermes Trader is an autonomous AI trading pipeline that requires no human intervention to operate. Every five minutes it ingests financial news from eleven sources, scores sentiment, and embeds signals into a searchable vector database. Every morning it generates sector predictions using a reasoning model. Three times daily it reviews the portfolio, entering and exiting positions based on strict rules. The system has learned that markets are not independent. Energy prices ripple into airline costs. Semiconductor supply chains move chip stocks before earnings. Fed language shifts entire sectors before policy is announced. These relationships are discovered through post-mortem analysis and stored as inference rules fed back into every future prediction. Market calendar context is also weighted — earnings seasons, Fed meeting cycles, and options expiration windows influence how signals are interpreted. Every prediction passes through six protection layers before executing: a recency filter, sector win-rate calibration, a meta-correction penalty, direction calibration, a hard-block threshold for underperforming sectors, and an independent critic AI that challenges every prediction. Seven portfolio gates then validate every trade — covering market hours, drawdown, position size, sector exposure, confidence, weekly limits, and volatility. When a trade fails, an automated post-mortem writes a new rule so the mistake won't repeat. Hermes is the natural language interface to the pipeline. It delivers a daily morning briefing, answers questions about positions and predictions in plain English, and validates or executes trades on request using the same safety checks the pipeline uses. Built on Alpaca Markets, locally-hosted open source models, and the Hermes Agent framework.