Which AI Hedge Fund GitHub Projects Are Worth Knowing About?

Updated 2026-09-25

TradingAgents and AI Hedge Fund (virattt) are two of the best-known open-source projects in this space. Both are open-source, both use LLMs to analyze stocks, and both come with disclaimers that are easy to skim past.

Here's what each project's own documentation actually claims, how they're built, and where a live public paper-trading experiment like this one sits next to them structurally.

TradingAgents: What Its README and Paper Say

TradingAgents (GitHub: TauricResearch/TradingAgents) is an Apache 2.0 licensed, open-source multi-agent framework built on LangGraph, according to its GitHub page. It produces a per-ticker research report and a buy/hold/sell decision.

Per its README, the pipeline runs in stages: an Analyst Team (fundamentals, sentiment, news, technical analysts) feeds a Researcher Team of bull and bear researchers who debate the findings, a Trader Agent drafts a decision, and a Risk Management team plus a Portfolio Manager review it before it would be executed on a simulated exchange.

The README says it supports many LLM providers behind each agent, including OpenAI, Google Gemini, Anthropic Claude, xAI Grok, DeepSeek, Qwen, GLM, MiniMax, OpenRouter, Mistral, Kimi, Groq, NVIDIA NIM, local Ollama models, Azure OpenAI, and AWS Bedrock. You can start a run from its command-line tool or as a Python call (TradingAgentsGraph().propagate(ticker, date)), and it ships a separate backtest utility that runs the same pipeline across a grid of tickers and dates.

Its own README is direct about a limitation: two runs on the same ticker and date can produce different output, because LLM sampling is non-deterministic and live news or social data changes between runs. It calls this expected for a research tool built on language models, not a defect, and notes backtest results may not match any previously published figure. Its disclaimer states the framework is designed for research purposes and is not intended as financial, investment, or trading advice. Its paper (arXiv 2412.20138) reports improvements in cumulative returns, Sharpe ratio, and maximum drawdown over baseline models — but that is the paper's own reported result in its own experiments, not an independent or live track record.

AI Hedge Fund (virattt): What Its README Says

AI Hedge Fund (GitHub: virattt/ai-hedge-fund) is an MIT-licensed open-source project. Its README describes it as a proof of concept exploring how AI could be used to make trading decisions, and it states plainly that it is for educational and research purposes only, is not intended for real trading, and that the system does not actually make any trades.

It installs as a command-line app called aihf. Per its README, you build a "fund" — a mandate file covering strategies, staff, risk, capital, and cadence — and point it at tickers. Its investor agents run on a model provider you choose from a list that includes Anthropic, OpenAI, DeepSeek, Google, xAI, Kimi, or TypeSafe, and it needs a Financial Datasets API key for prices, fundamentals, and earnings. The README also says a saved fund can be backtested over history at its rebalance cadence.

How the Two Projects Compare

Star counts are popularity, not a performance signal — neither project's README or paper claims otherwise, and neither points to a live, forward track record.

Where a Live Paper-Trading Experiment Fits Differently

This site runs about twenty trading bots on live paper accounts (simulated money, real market prices), with rulebooks rewritten every day or every third day by AI models from the GPT, Claude, Grok, and Gemini families, plus a fixed-rules System and a few frozen rulebooks. Every bot is ranked against the S&P 500 and every trade is published; one bot's trades are also mirrored to a small real-money account, shown factually on the site.

The structural difference from TradingAgents is worth naming plainly: TradingAgents makes fresh LLM calls at every step of every run, which is why, by its own README, identical inputs can produce different output, and why it can't be replayed and scored in the usual fixed-rulebook sense. A rulebook here, once an AI has written it for the day, is deterministic and mechanical for that day, so it can be backtested or replayed exactly. That's a difference in design, not a claim that one approach beats the other — someone who wants a per-ticker research report with bull/bear debate may well prefer TradingAgents to this site, and that's a reasonable choice.

See the live paper-trading scoreboard — free — stocks

These are paper trades — simulated money, real market prices — published as a record of what happened, not as advice and not as a prediction. Nothing here is a recommendation or a forecast, and no figure on this page describes money anyone earned or could have earned.