Stock Wars (github.com)

A new autonomous stock trading agent framework has been announced, incorporating Large Language Models (LLMs) and tools like OpenBB and FMP for seamless, real-time trading. The system utilizes a ReAct (Reasoning + Acting) loop to process market data effectively, facilitating both robust backtesting and live trading capabilities. This framework empowers traders to leverage AI’s analytical strengths by employing models like GPT-5 and DeepSeek to interpret technical and fundamental data, integrated with high-quality financial information from OpenBB and FMP. Significantly, this development represents a leap forward in algorithmic trading, allowing users to perform extensive historical simulations and real-time trades with improved reliability through features like a PM2-managed live trading daemon. The system supports multi-stock analysis and provides detailed portfolio management, tracking cash and positions while calculating profit and loss dynamically. With the added functionality for parallel processing and configurable trade allocation, this framework not only enhances trading efficiency but also sets the stage for more sophisticated AI strategies in financial markets, addressing issues like lookahead bias in data analysis.