Tauric Research has released TradingAgents v0.4.0, an open-source multi-agent LLM framework designed for financial trading research and deployment. The framework uses a team of specialized AI agents — each handling different aspects of the trading pipeline — to analyze markets, generate signals, and manage portfolios.
The v0.4.0 release introduces look-ahead and point-in-time fixes across FRED macro data, social sentiment analysis, and the decision-log memory system. It also adds clearer decision signals, a working CLI checkpoint resume feature, and Trader price grounding to reduce hallucination in financial contexts. The framework supports current-generation models including GPT-5.6 and GLM-5.3.
TradingAgents is built on LangGraph and supports a wide provider registry: NVIDIA, Kimi, Groq, Mistral, Bedrock, and any OpenAI-compatible endpoint. The architecture separates concerns across Research Manager, Trader, and Portfolio Manager agents, each with structured-output capabilities and persistent decision logging.
The project addresses a real gap in AI-for-finance tooling: most LLM trading demos are single-agent systems that conflate analysis, execution, and risk management into one prompt. TradingAgents' multi-agent design mirrors how actual trading desks operate — with specialists handling research, execution, and portfolio construction separately. The look-ahead-safe data pipeline is particularly important for backtesting integrity, preventing the model from accidentally using future information in historical simulations.