TauricResearch/TradingAgents
A framework where multiple LLM agents split into analyst, researcher, trader and risk-management roles, debate, and decide stock trades
TradingAgents mimics the structure of a real trading firm, using LLM-powered agents for fundamentals, sentiment, news and technical analysis, bullish/bearish researchers who debate, a trader who synthesizes their reports, and risk-management plus portfolio-manager agents who approve final trades. It's built with LangGraph to wire the agents' workflow together and supports many LLM providers including OpenAI, Google, Anthropic, xAI, DeepSeek, Qwen, GLM, MiniMax, Ollama for local models, and Azure. It works on any ticker covered by Yahoo Finance across markets like the US, Hong Kong, Tokyo, London, India, China A-shares and crypto, and it logs past decisions to inform future runs plus supports checkpoint resume for interrupted sessions.
What it does
- Analyst team (fundamentals, sentiment, news, technical indicators), researcher team (bull vs bear debate), trader agent, and risk-management/portfolio-manager agents collaborate in sequence
- Built with LangGraph to structure the agent workflow as a graph, and designed to swap between many different LLM provider APIs
- Completed trading decisions are logged locally and reused as reflection material when analyzing the same ticker again later
- The README explicitly notes that running the same ticker and date twice can produce different results due to LLM sampling randomness and live news/social data changing over time
- Positioned explicitly as a research tool, not a strategy guaranteeing any fixed, replicable return
Why it matters
It offers a concrete, runnable example of AI agents with different perspectives (finance, news, technicals, risk) debating to reach a decision, which is useful reference material for anyone designing multi-agent LLM systems or researching automated financial analysis.
Terms in this repo
- LLM · Large language model, an AI model that understands and generates text
- LangGraph · A framework for structuring multiple AI agents' workflow as a graph
- Checkpoint resume · Saving intermediate state so an interrupted run can continue instead of restarting
- MACD, RSI · Common technical indicators used to analyze price trends
- Alpha · Financial term for return in excess of a market benchmark such as SPY
Repository description (English)
TradingAgents: Multi-Agents LLM Financial Trading Framework
Open on GitHubTrending repos
- DietrichGebert/ponytailA ruleset that teaches AI coding agents to say no to writing code
- liustack/modlensA plugin that lets text-only AI models like DeepSeek 'see' pasted images by turning them into structured text data
- marin-community/marinMarin is an open-source platform that publishes the entire process of building large language models, not just the final model
- Shubhamsaboo/awesome-llm-appsA free collection of 100+ ready-to-run AI agent and RAG app examples
- asciimoo/histerA personal search engine that indexes the full text of every page and file you've seen
- anthropics/claude-plugins-officialAnthropic's official catalog for add-ons that plug new abilities into Claude Code
- vorssaint/vorssaint-utilsOne free menu bar app replaces a dozen paid Mac utilities
- MadsLorentzen/ai-job-searchAn open-source framework that runs your entire job search through Claude Code, built by someone who actually got hired using it
Latest from METAL LAB
- Anthropic funds $5M research program for AI wellbeing evaluations
- Keenable Raises $26M Seed to Build Web Search for AI Agents
- Gamma acquires Accel-backed design startup Lica
- Oura's rumored September IPO is being priced on AI, not the ring
- Alabama Launches Investigation into OpenAI Agent Breakout Incident