TradingAgents: 12 LLM agents mimic a trading firm's workflow

Tauric Research's open-source Python framework TradingAgents, built on LangGraph, runs four analysts, a bull-bear debate, a trader, three risk analysts and a portfolio manager to rate a stock. Version 0.7.0, released October 16, 2026, makes at least 15 model calls per decision, 13 of them on the quick model.

• The GitHub project has 115,371 stars and over 22,000 forks
• One decision takes at least 15 LLM calls: 13 quick and 2 deep
• Defaults are OpenAI's GPT-6 Luna and GPT-6 Sol
• No broker integration: output is one of five ratings from Buy to Sell