Markets generate more data in a single day than any human could process in a lifetime — price action, earnings transcripts, SEC filings, macro indicators, and breaking news all happening at once. Traditional technical analysis and manual research simply can’t keep up. AI can. It scans thousands of stocks simultaneously, backtests decades of data in seconds, and can even simulate entire trading desks where multiple agents debate whether a trade is worth taking.
The best part? You don’t need a hedge fund budget anymore. The open-source community on GitHub has built tools that give retail traders, developers, and finance students access to the same categories of AI-powered analysis that institutions use
We’ve ranked the 10 best AI trading repositories on GitHub by stars. All are open source. All are free to run. None of this is financial advice.
1. 🏆 TradingAgents — ~108.7k ⭐
Repo: TauricResearch/TradingAgents · Type: Research
TradingAgents is the runaway leader and for good reason. It simulates an entire AI-powered trading desk where multiple LLM agents debate the merits of a trade before executing. You get bull analysts, bear analysts, a risk manager, and a trader — all arguing with each other in real time. It’s like watching a Bloomberg terminal come alive.
Best for: Seeing how an AI desk debates a trade.
2. ai-hedge-fund — ~63.8k ⭐
Repo: virattt/ai-hedge-fund · Type: Paper Trading
What if Warren Buffett, Ray Dalio, and Cathie Wood were all AI agents arguing about whether to buy Tesla? That’s essentially what ai-hedge-fund does. Each agent embodies the investment philosophy of a legendary investor, and a portfolio manager agent synthesizes their arguments into a final decision. It’s brilliant for learning how multi-agent systems can model diverse investment styles.
Best for: Learning multi-agent investing.
3. Vibe-Trading — ~34.1k ⭐
Repo: HKUDS/Vibe-Trading · Type: Paper Trading
Vibe-Trading brings the “vibe coding” ethos to the markets. Type what you want in plain English — “Show me momentum stocks that broke out above their 50-day MA this week” — and get backtests, chart analysis, and trade signals. It’s the most accessible entry point for retail traders who want an AI copilot without writing a single line of code.
Best for: Retail traders who want a CLI copilot.
4. FinGPT — ~21.3k ⭐
Repo: AI4Finance-Foundation/FinGPT · Type: Research
FinGPT is the foundational layer for anyone who wants to build their own financial AI model. It provides open-source financial LLMs alongside fine-tuning recipes tailored for sentiment analysis, SEC filings, and financial news. Think of it as the “Hugging Face for finance” — a toolkit that lets you train models that actually understand money.
Best for: Building your own finance model.
5. AI-Trader — ~21.0k ⭐
Repo: HKUDS/AI-Trader · Type: Paper Trading
AI-Trader turns trading into a competitive sport. You spin up multiple trading agents, each with its own strategy, and let them trade in parallel. They share what they learn, rank on a leaderboard, and you can watch Darwin’s theory play out in your portfolio. It’s gamification meets quantitative finance.
Best for: Pitting agents against each other.
6. FinRL — ~16.5k ⭐
Repo: AI4Finance-Foundation/FinRL · Type: Research
FinRL is the gold standard for applying reinforcement learning to financial markets. It provides a complete pipeline — from training RL agents on historical data, to backtesting strategies, to deployment. If you’re a quant who believes that the market is a game and RL is the way to win it, this is your starting point.
Best for: Quants training RL policies.
7. ai-berkshire — ~13.2k ⭐
Repo: xbt1in/ai-berkshire · Type: Research
This one is for the Warren Buffett disciples. ai-berkshire is a plugin for coding agents (Claude Code, Codex) that runs structured value-investing analysis — DCF models, comparable company analysis, and intrinsic value calculations. Instead of a flashy trading bot, it’s a thoughtful research assistant that helps you do deep fundamental analysis right inside your IDE.
Best for: Buffett-style research in your coding agent.
8. nofx — ~13.0k ⭐
Repo: NoFxAIOS/nofx · Type: 🔴 LIVE
This is the only repo on the list built for live trading. nofx is an open-source AI trading terminal that takes risk management seriously — it enforces position limits, drawdown limits, and includes a physical halt switch in code. If you’re going to let an AI trade real money, nofx is the most responsible way to do it.
CAUTION
nofx is the only live-trading tool on this list. All others are research or paper-trade only. Paper trade first, always. Real money means real risk.
Best for: The one built for live trading.
9. FinRobot — ~8.1k ⭐
Repo: AI4Finance-Foundation/FinRobot · Type: Research
FinRobot automates what junior equity analysts spend 80 hours a week doing: reading SEC filings, listening to earnings calls, and distilling it all into structured research reports. Feed it a ticker symbol and it produces a report that covers financials, management commentary, risk factors, and a thesis — all sourced and cited.
Best for: Automated equity research.
10. OpenAlice — ~7.2k ⭐
Repo: TraderAlice/OpenAlice · Type: Approval-Gated
OpenAlice takes the opposite approach to full automation. It’s a trading workspace designed for coding agents, but every single action that touches your brokerage account requires explicit human approval. Nothing happens without you clicking “yes.” For anyone who wants AI assistance but isn’t ready to hand over the keys, OpenAlice is your safety net.
Best for: Keeping a human in the loop.
Key Takeaways
Live vs. Research vs. Paper
| Category | Count | What It Means |
|---|---|---|
| Research | 5 repos | Academic / experimental — not meant for real trades |
| Paper | 3 repos | Simulated trades with no real money at risk |
| Live | 1 repo (nofx) | Built for real-money trading with risk controls |
| Approval-Gated | 1 repo (OpenAlice) | AI suggests, human approves every action |
IMPORTANT
Backtests aren’t returns. One TradingAgents 30-day test beat the S&P (~7% vs 4.5%) with a 22% drawdown. Paper trade first, always.
Where to Start
- New to AI trading? Start with TradingAgents — it’s the most popular for a reason and gives you a mental model of how an AI trading desk operates.
- Want to go live? nofx is the only responsible option — but only after extensive paper trading.
- Building something custom? FinGPT + FinRL give you the foundational models and RL framework to build from scratch.
- Just want research? FinRobot will save you dozens of hours reading filings and earnings transcripts.
Final Thoughts
The AI trading landscape on GitHub is maturing fast. What started as toy projects and academic papers has evolved into sophisticated multi-agent systems, reinforcement learning pipelines, and even live trading terminals with real risk management. The barrier to entry has never been lower.
But a word of caution: these tools are powerful, not profitable by default. A backtest that beats the S&P by 2.5% over 30 days with a 22% drawdown is not a money printer — it’s a research signal. Treat these repos as learning tools first, paper-trading tools second, and live-trading tools only after extensive validation.
The future of trading is undeniably AI-augmented. These 10 repos are the best starting points on the internet today.
Get real time update about this category directly on your device, subscribe now.