Overview

TradingAgents is a cutting-edge, multi-agent framework that harnesses the power of Large Language Models (LLMs) to simulate the complex dynamics of financial trading firms. It deploys a team of specialized LLM-powered agents—from fundamental, sentiment, and technical analysts to traders, risk managers, and portfolio managers—to collaboratively analyze market conditions, engage in dynamic discussions, and ultimately inform strategic trading decisions. Designed for research purposes, TradingAgents offers a sophisticated platform for exploring AI-driven financial analysis and strategy development.

Key Features

• Multi-Agent Architecture: Mimics real-world trading firms by assigning distinct roles and expertise to specialized LLM agents, including Analysts (Fundamentals, Sentiment, News, Technical), Researchers, Traders, and Risk Management/Portfolio Managers.
• Dynamic Agent Collaboration: Agents engage in structured debates and discussions to critically assess information, balance potential gains against risks, and refine trading strategies.
• Extensive LLM Provider Support: Integrates with a wide array of LLM providers including OpenAI (GPT series), Google (Gemini), Anthropic (Claude), xAI (Grok), DeepSeek, Qwen, GLM, MiniMax, OpenRouter, and local models via Ollama, as well as enterprise solutions like Azure OpenAI and AWS Bedrock.
• Flexible Configuration: Highly customizable through a detailed configuration system, allowing users to select LLM providers, specify models for different tasks (deep vs. quick thinking), adjust debate rounds, and set sampling temperatures.
• Robust Data Integration: Supports various data sources, including financial data from Yahoo Finance (covering global markets, equities, and crypto) and economic data from FRED.
• Persistence and Recovery: Features a persistent decision log that carries forward insights and lessons learned from previous trading sessions, enhancing the learning process. It also offers optional checkpoint resuming for interrupted runs, ensuring progress is not lost.
• Reproducibility Controls: Provides mechanisms to manage and understand variation in LLM-driven outputs, including deterministic company identity resolution and grounded data snapshots, while acknowledging the inherent non-determinism of LLMs.
• Intuitive CLI and Python API: Offers both an interactive command-line interface (CLI) for easy exploration and a Python package for seamless integration into custom applications and research workflows.

Typical Use Cases

• AI-Driven Financial Research: Investigate how LLM-powered agents can analyze market data, news, and sentiment to generate trading insights and identify potential investment opportunities.
• Algorithmic Trading Strategy Development: Explore the creation and refinement of trading strategies by leveraging the collective intelligence and specialized functions of multiple AI agents.
• Agent-Based Financial Modeling: Study the emergent behaviors and decision-making processes of complex multi-agent systems in simulated financial market environments.
• LLM Application in Finance: Serve as a foundational research scaffold for developing and testing advanced LLM applications within the financial domain, beyond simple prediction tasks.
• Understanding Market Dynamics: Gain deeper insights into how different facets of financial analysis—from fundamentals to sentiment—interact and influence trading decisions.