TradingAgents is a high-performance, open-source multi-agent framework designed to simulate the decision-making dynamics of a professional financial trading firm. Built using LangGraph, it orchestrates a team of specialized LLM-powered agents—including fundamental, technical, and sentiment analysts—who collaborate and debate to evaluate market conditions. The framework is designed for research and backtesting, supporting a wide array of state-of-the-art models like GPT-5.4, Claude 4.6, and Gemini 3.1 to inform automated trading decisions.
Use Cases
Automated Market Sentiment Analysis
Deploy specialized agents to scrape and score social media and news feeds, gauging the short-term market ‘mood’ for specific tickers.
Fundamental & Technical Backtesting
Analyze historical stock performance by combining company financials with technical indicators (MACD, RSI) to validate trading strategies.
Multi-Agent Debate & Strategy Refinement
Simulate ‘bullish’ and ‘bearish’ researcher personas to engage in structured debates, balancing potential gains against inherent risks before a trade is proposed.
Risk-Adjusted Portfolio Management
Utilize a dedicated Risk Management agent to evaluate market volatility and liquidity, providing a final ‘approve/reject’ gate for transaction proposals.
LLM Benchmark for Finance
Compare the reasoning and trading performance of different frontier models (e.g., GPT-5 vs. Claude 4) within a standardized financial environment.
Features & Benefits
Modular Analyst Team
Decomposes trading into specialized roles: Fundamentals Analyst, Sentiment Analyst, News Analyst, and Technical Analyst.
LangGraph Orchestration
Uses a graph-based architecture to ensure flexible, reliable, and non-linear communication between the different agent teams.
Multi-Provider LLM Support
Native integration for OpenAI, Google (Gemini), Anthropic (Claude), xAI (Grok), and OpenRouter, plus local support via Ollama.
Deep-Think vs. Quick-Think Logic
Configurable model settings allow for expensive ‘reasoning’ models for strategy and cheaper ‘fast’ models for data processing.
Interactive CLI & Reporting
A built-in command-line interface that visualizes agent progress in real-time and outputs comprehensive trading decision reports.
100% Free & Open Source
Released under the Apache-2.0 license, allowing developers and researchers to modify and extend the code for private or commercial use.
High Academic Fidelity
Based on a technical research report, lending a level of theoretical rigor to the agent interactions.
Local-First Potential
Support for Ollama allows users to run the entire trading framework locally for maximum data privacy and reduced API costs.
Cons
Research Only
Explicitly stated as a research tool; users must be cautious when applying these strategies to live capital due to model non-determinism.
Technical Complexity
Requires a solid understanding of Python, LangGraph, and financial data APIs (like Alpha Vantage) to configure effectively.
