Trading: 8 free AI trading repos that make f*ck you money

Watch on YouTube: https://www.youtube.com/watch?v=yuckuVqd01k

Channel: Miles Deutscher Finance and AI Edge

Eight Open-Source AI Trading Repositories Worth Your Attention

Open-source AI trading tools have matured rapidly. Miles Deutscher recently reviewed eight GitHub repositories that collectively hold roughly 400,000 stars, ranging from full hedge-fund simulations to lightweight daily report generators. All are free to clone, and the creator provides a School community with drag-and-drop setup prompts that let you install every repo in minutes using Claude or GPT. Below is a structured breakdown of what each project does, the key numbers behind them, and practical steps to start testing them yourself.

What the Video Covers

The walkthrough ranks the eight repos by GitHub stars and perceived utility, assigning tier grades from S to C. The top pick, Trading Agents (117,000 stars), simulates a 12-agent hedge-fund workflow — analysts, researchers, traders, risk managers, and a portfolio manager — with optional Alpaca execution for paper or live trading. Other highlights include:

• Daily Stock Analysis (67,000 stars): Automated daily Telegram reports with buy/watch/sell ratings, a 0–100 score, price levels, and risk alerts.
• AI Hedge Fund (64,000 stars): Persona-based analysis mimicking Buffett, Munger, Burry, and Wood for multi-perspective research.
• Freqtrade (56,000 stars, 349 contributors): Battle-tested crypto bot framework with dry-run mode, backtesting, ML strategy optimization, and Telegram control.
• Anthropic Financial Services Toolkit (40,800 stars): Specialized agents for investment banking, equity research, private equity, and wealth management; requires paid data (FactSet/Quiver Quant) for real edge.
• Vibe Trading (36,800 stars): Natural-language research, realistic-fee backtesting, and paper trading via Alpaca; includes a self-improving agent and multi-agent teams.
• Day Trader Skill (new, 11 Claude skills): Chart scanning, news monitoring, backtest replay, risk management, and order execution — still unproven.
• LLM Trading Lab (Tier C): Originated as a $100 real-money GPT experiment that returned 24% in one month, but lacks a repeatable, fully automated framework.

Step-by-Step Walkthrough

If you want to move from overview to hands-on testing, follow this sequence:

1. Join the free School community (link in the video description) and download the asset-library prompts.
2. Paste the master prompt into Claude or GPT to auto-install all eight repositories in your local environment or a cloud VM.
3. Start with Trading Agents (S-tier). Configure Alpaca paper-trading credentials, define a watchlist, and run the 12-agent loop to see research, risk, and execution outputs.
4. Enable Daily Stock Analysis for your core watchlist (e.g., NVDA, AAPL, TSLA, MSFT). The Telegram bot delivers a concise report each morning at near-zero marginal cost.
5. Experiment with Vibe Trading: ask it in plain English to “find and backtest a momentum strategy on mega-cap tech,” then paper-trade the resulting signals with the default $100,000 virtual capital.
6. Run Freqtrade in dry-run mode on BTC/ETH pairs. Use the built-in ML optimizer to iterate on strategy parameters before risking real capital.
7. Query AI Hedge Fund before a discretionary entry to collect Buffett-style value, Munger-style quality, Burry-style contrarian, and Wood-style growth perspectives.
8. Combine your favorite modules — e.g., Trading Agents research + Freqtrade execution + investor personas — into a custom repository and build a simple dashboard (Streamlit, Dash, or a Next.js frontend) to visualize signals, P&L, and risk metrics.

Key Numbers & Tools at a Glance

• Combined GitHub stars: ~400,000
• Trading Agents: 117k stars, 12 agents, Alpaca execution support
• Daily Stock Analysis: 67k stars, Telegram delivery, sub-cent cost per report
• AI Hedge Fund: 64k stars, four famous-investor personas
• Freqtrade: 56k stars, 349 contributors, since 2017
• Anthropic Financial Services Toolkit: 40.8k stars, requires FactSet/Quiver Quant subscriptions
• Vibe Trading: 36.8k stars, $100k paper capital default, university-lab origin
• Day Trader Skill: 11 Claude skills, no live track record yet
• LLM Trading Lab: $100 real-money test, 24% in 30 days (single run)

Caveats & Hype Check

“Closest thing to having your own research team” — but Trading Agents shows no verified live P&L.

“Like having a hedge fund in your pocket” — yet the AI Hedge Fund repo admits “elements of gimmick” and delivers generalized, not real-time, insights.

• No live performance data for the flagship repos; all results are backtests or paper-trading logs.
• Strategy dependency: Freqtrade and Vibe Trading are only as good as the trading logic you supply.
• Data costs: Anthropic’s toolkit needs paid datasets (FactSet, Quiver Quant) to move beyond demo mode.
• Single-experiment claims: LLM Trading Lab’s 24% return on $100 over one month is not repeatable evidence.
• Lead funnel: The free School community is a gateway to the creator’s broader ecosystem; no affiliate links are pushed, but the funnel exists.

Concrete Next Actions

• Join the School community and grab the install prompts today.
• Run Trading Agents in paper mode for at least two weeks; log every signal, fill, and risk decision.
• Activate Daily Stock Analysis for your watchlist and compare its scores against your own thesis.
• Backtest a crypto strategy in Freqtrade dry-run with ML optimization enabled; export the parameter set.
• Prompt Vibe Trading to discover a mean-reversion setup on SPY/QQQ, then paper-trade it for a month.
• Run AI Hedge Fund on your next three potential buys; document where personas agree or diverge.
• Fork the repos you like, stitch the research layer (Trading Agents) to the execution layer (Freqtrade/Alpaca), and add a lightweight UI dashboard.
• Consider open-sourcing your combined system — contribution back to the community accelerates everyone’s learning curve.

These repositories lower the barrier to agentic finance, but they are tools, not turnkey money printers. Treat them as programmable building blocks: validate each component in isolation, measure real paper-trading metrics, and only then allocate risk capital. The fastest path to a durable edge is iterating on your own strategy logic inside these open frameworks — not chasing someone else’s backtest.