The gist

TradingAgents is an open-source, multi-agent framework by Tauric Research for simulating financial trading. It uses specialized LLM-powered agents—such as analysts, researchers, and portfolio managers—to collaboratively evaluate market conditions, debate strategies, and inform trading decisions. The framework is designed to decompose complex trading tasks into distinct roles, enabling a more robust approach to market analysis research. It is built using LangGraph for modularity.

What it does

Deploys specialized agents for roles like fundamental, sentiment, and technical analysis.
Orchestrates structured debates between agents to balance risks and potential gains.
Generates trading decisions based on comprehensive insights from analyst and researcher agents.
Simulates risk management and portfolio approval workflows.
Maintains a persistent decision log to learn from past performance.
Integrates with multiple LLM providers, including OpenAI, Google, Anthropic, and local models via Ollama.

How it works

TradingAgents is an open-source Python package run via a command-line interface or imported as a module. A user provides a stock ticker and analysis date. The framework then orchestrates a team of LLM agents to research the stock, debate findings, and propose a trade. The final output is a trading decision. The framework is free to use but requires user-provided API keys for LLM access.

Best for

This framework is best for quantitative researchers and AI developers looking to experiment with and backtest multi-agent systems for financial market analysis and automated trading strategies.

Watch out for

The framework is intended strictly for research purposes and is not financial advice. Trading performance is variable and depends heavily on the LLMs, data quality, and other non-deterministic factors.