The article discusses the development of a self-improving AI trading department using RAG-powered agents. The system includes Fundamental Analyst, Technical Analyst, Sentiment Analyst, and Manager Agent. Each agent specializes in a specific domain like fundamental analysis, technical indicators, and sentiment analysis. The Manager Agent consolidates all decisions to make the final call. The system continuously learns from past mistakes and aims to build a smarter decision engine over time. Future enhancements may include a Risk Manager Agent, Portfolio Allocator Agent, and Quantitative Evaluator. The project is open for collaboration and further development. The full code is available on Kaggle for exploration and contributions.