TradingAgents, a multi-agent LLM financial trading framework by TauricResearch, released version 0.4.0 featuring look-ahead and point-in-time fixes across FRED macro data, social sentiment, and decision-log memory. The update includes clearer decision signals, working CLI checkpoint resume, Trader price grounding, and support for GPT-5.6 and GLM-5.3 models. Previous v0.3.1 brought Alpha Vantage look-ahead filtering, graph-router crash-safety, and Claude Sonnet 5 support.

Version 0.3.0 expanded the provider registry to include NVIDIA, Kimi, Groq, Mistral, Bedrock, and any OpenAI-compatible endpoint, plus FRED and Polymarket data vendors. The framework mirrors real-world trading firms with specialized agents: Fundamentals, Sentiment, News, and Technical Analysts; bullish and bearish Researchers; a Trader Agent; and Risk Management. Agents engage in dynamic discussions to determine optimal strategies.

Designed for research, TradingAgents performance varies by backbone models, temperature, data quality, and non-deterministic factors. It is not financial advice. Installation supports Docker, remote Ollama, and TRADINGAGENTS_* environment variables with API-key auto-detection.