TradingAgents: Simulate a trading firm's AI team analyzing a stock

What it does

TradingAgents copies how a trading firm works using AI agents, each a language model given one job. Analysts study company finances, news, social media mood and price charts at the same time. Researchers then argue for and against using their reports, a trader proposes a trade, a risk team reviews it, and a portfolio manager approves or rejects it.

How to use it

Clone the project, create a Python 3.11+ environment and install the package; or copy the example settings file, add your API keys (access codes), and run it with Docker Compose. Set the key for your chosen provider, or pick Subscription in the provider menu to use a ChatGPT plan. Then run the command-line tool with a ticker (stock symbol) and date.

Common use cases

• Give the command-line tool a stock symbol and a date, and the analysts, researchers and the trader produce a decision saved as one HTML report page.
• Run a backtest over several tickers and dates, and you get past decisions scored from the market closes that followed, using only data published by each date.
• Include the Sentiment Analyst for a ticker, and you get news headlines, StockTwits and Reddit chatter summed up as one short-term market-mood reading.