TL;DR
TradingAgents is an open-source framework where multiple LLM agents collaborate to make financial trading decisions.
Key features
Multi-agent architecture: Multiple agents such as research manager, trader, and portfolio manager divide roles and collaborate.
Support for various LLM providers: Supports multiple models including GPT, Gemini, Claude, Grok, DeepSeek, Qwen, GLM, etc.
Backtesting and real-time trading: Test strategies with historical data and apply to real-time markets.
Structured output: Each agent's decisions are output in a structured format for transparency.
Checkpoint and resume: Based on LangGraph, interrupted workflows can be resumed.
When to use it
When you want to automate financial trading strategies with LLM-based agents.
When you want to compare and evaluate the trading performance of different LLM models.
When researching or prototyping financial applications of multi-agent systems.