Multi-agent trading system
Production-grade multi-agent swing trading bot with adaptive learning, multi-timeframe confirmation and comprehensive risk management — two independent pipelines: a LangGraph stock flow and a leveraged-ETF specialist.
A swing trade is usually one person's gut feel on one timeframe — entry, exit and position size decided by a single model nobody can audit — and leveraged ETFs decay in ways a stock playbook quietly ignores.
Two independent pipelines instead of one do-everything model. The stock pipeline is a LangGraph multi-agent flow: a regime classifier sets the market context, a scanner shortlists candidates, then fundamental and sentiment agents run in parallel and join into a signal combiner with SHAP-explained features and a confidence gate, before a risk manager sizes the trade and an executor places it. The ETF pipeline is a specialist for leveraged funds: regime and relative-strength rotation, volatility-targeted sizing, decay-aware stops and a portfolio circuit breaker, with an options, news and intraday layer that can be switched on per setup.
A system that reads the market, forms a thesis, sizes the risk and writes down why — every stage independently backtested, and every new idea gated behind its own confidence check. Scheduled agents review results weekly and learn after every trade, so the evidence trail keeps growing.
Want to talk through this project?