TauricResearch/TradingAgents

Traditional trading systems rely on single-model predictions or rule-based strategies, which lack the nuanced, multi-perspective analysis and risk management of human trading desks. TradingAgents solves this by orchestrating multiple specialized LLM agents that debate and reach consensus, providing a more holistic and robust decision-making framework. It also addresses the complexity of integrating multiple data sources (fundamental, sentiment, technical) and LLM providers into a cohesive pipeline.

Developers can build fully automated trading bots, backtesting systems, and decision-support tools that analyze stocks, cryptocurrencies, and other assets using a multi-agent debate framework. The framework supports multiple LLM providers (OpenAI, Anthropic, Google, etc.), data vendors (Alpha Vantage, FRED, Polymarket), and includes features like structured outputs, checkpointing, and Docker deployment. The ceiling includes deploying sophisticated trading strategies that incorporate fundamental, sentiment, and technical analysis with risk management, and even extending to new asset classes or custom agent roles.