TradingAgents: Open-Source Multi-Agent LLM Framework for Financial Trading and Market Analysis

TradingAgents is an open-source Python framework implementing multi-agent LLM architecture for financial trading and market analysis. The framework coordinates specialized agents responsible for different aspects of the trading workflow: market data analysis agents that process news, earnings, and macroeconomic indicators; strategy agents that generate and debate trading hypotheses; portfolio management agents that allocate capital across positions; risk assessment agents that evaluate downside scenarios and position sizing; and execution agents that interface with brokerage APIs. The multi-agent design enables both collaborative decision-making where agents share findings and competitive filtering where conflicting signals trigger debate before final position sizing. Built entirely on open-source LLMs, the framework allows firms to keep trading data on-premise while customizing agent behaviors for specific asset classes, geographies, or regulatory requirements. Backtesting modules evaluate agent performance against historical data with configurable slippage and fee models. TradingAgents targets quantitative research teams, hedge funds, and retail trading platforms seeking to experiment with LLM-driven trading strategies without proprietary vendor lock-in or data leaving their infrastructure.