Multi-Agent LLM-Based Trading Framework 'Trading Agents' Released

Trading Agents is a multi-agent LLM framework that mimics the structure of an actual Wall Street trading firm. It is built in Python and has gained significant attention, recording over 53,000 stars on GitHub.

The system consists of agents that perform the following specialized roles:

• Analyst Team: Fundamental, sentiment, news, and technical analysis agents work in parallel to gather data.
• Research Team: Bull and Bear agents debate with each other to derive the best conclusion.
• Trading Team: Traders who decide the timing and size of trades, a risk management team that assesses risk, and a portfolio manager responsible for final approval work together.

Technically, it uses LangGraph to manage transitions between agents as nodes, and the key point is that the entire decision-making process is logged, enabling traceability and auditability. Unlike existing black-box machine learning models, this provides clear logical grounds for why a particular trade was proposed or rejected.