TradingAgents
Multi-agent LLM framework that simulates a trading firm with specialized AI analysts
trading-agents multi-agent llm open-source langraph algorithmic-trading risk-management tauric-research
OVERVIEW
TradingAgents is an open-source multi-agent trading framework developed by Tauric Research that mirrors the structure of a real-world trading firm. It deploys specialized LLM-powered agents — including fundamental analysts, sentiment analysts, technical analysts, news analysts, traders, and risk managers — that collaboratively evaluate market conditions and make trading decisions.
The framework supports multiple LLM providers including GPT-5, Claude 4, Gemini 3, and Grok 4, selecting models based on task complexity. Quick-thinking models handle data retrieval while deep-thinking models perform in-depth analysis. Built on LangGraph for modularity, TradingAgents requires no GPUs to run.
Research published at arXiv demonstrates its superiority over baseline models with notable improvements in cumulative returns, Sharpe ratio, and maximum drawdown metrics. Version 0.2.0 (February 2026) introduced multi-provider support and improved architecture.
ADVANTAGES
- + Simulates a full trading firm with 7 specialized agent roles
- + Supports multiple LLM providers (GPT-5, Claude 4, Gemini 3, Grok 4)
- + No GPU required — runs entirely on LLM API calls
- + Open-source with active development and academic research backing
- + Improved Sharpe ratio and drawdown vs baselines
- + Built on LangGraph for flexibility and modularity
LIMITATIONS
- - Requires LLM API keys (costs depend on provider usage)
- - No web interface — command-line and code only
- - Relatively new project (v0.2.0 as of Feb 2026)
- - Limited to stock trading currently