Open-source quant tools by job: data, backtest, indicators, performance, portfolio, agents
Name the job first. Then install one tool for that job.
Short answer: a quant stack is several jobs, not one product: get data, test a strategy, compute indicators, judge the result, and split the money. You do not install all of them on day one. Pick one data library and one backtester, run one simple strategy, and add the rest only when that loop works. If you are still choosing the order, start with how to choose AI quant tools.
Each group below lists a few widely used open-source projects. The one-line descriptions were checked against the projects' own READMEs on 13 Oct 2026. Stars and maintenance status are read from GitHub and refresh with the daily snapshot. This site has no commercial relationship with these projects. The same list in Chinese is the Chinese tool list.
01 · Data: get prices first
Everything later depends on data. For personal research, start with a free library. Move to a paid source when that is not stable enough.
| Tool | What it is | GitHub stars | Maintenance |
|---|---|---|---|
| akshare | Free library that aggregates data from public finance sites, mainly Chinese markets and macro data | 23.1k | Last commit 4 days ago |
| yfinance | Free library that reads Yahoo Finance's public endpoints, suited to US stock research | 25.6k | Commits today |
| OpenBB | Open-source connector that attaches data vendors to one interface. It does not produce data itself | 74.8k | Last commit 3 days ago |
How to choose among the data libraries, including the paid EODHD APIs service, is the yfinance, AKShare, OpenBB, and EODHD guide. A worked daily update into SQLite is the daily price pipeline. The hands-on AKShare walkthrough for China A-shares is AKShare into SQLite. The same walkthrough in Chinese is in Chinese. Putting EODHD daily bars on that same table is the EODHD walkthrough. That walkthrough is not the four-library comparison above.
02 · Backtesting and trading frameworks: test a strategy
Run a strategy on history. Frameworks differ mainly in speed, which markets they cover, and whether they can send live orders.
| Tool | What it is | GitHub stars | Maintenance |
|---|---|---|---|
| backtrader | Backtesting and live-trading platform written in Python. The repository has been quiet for a long time | 23.6k | No commits in over six months |
| vnpy | Open-source quant trading system framework written in Python (VeighNa) | 46.1k | Commits today |
| Qlib | Microsoft's open-source AI quant platform, covering data processing, model training, and backtesting | 49.7k | Last commit 5 days ago |
| nautilus_trader | Algorithmic trading engine with a Rust core. Strategies are written in Python. Backtest and live trading share one codebase | 30.9k | Commits today |
| vectorbt | Vectorized backtesting and strategy research toolkit built on pandas, NumPy, and Numba. There is a separate paid PRO edition | 9.6k | Last commit 4 days ago |
| Lean | QuantConnect's open-source algorithmic trading engine, for backtests and live trading, with strategies in Python or C# | 22.3k | Commits today |
backtrader still runs, and its repository has been quiet for a long time, so a new project you expect to keep is better on a framework that is still maintained. The nine-framework comparison, including vnpy, RQAlpha, Qlib, and vectorbt, is how to choose a backtesting framework. A shorter note on backtrader alone is backtrader and four alternatives. A worked A-share example is the first backtrader strategy. The same article in Chinese is the Chinese edition. The Qlib install and first Alpha158 run are the Qlib walkthrough. More frameworks are in the backtesting category.
03 · Technical indicators: moving averages, MACD, and the rest
You do not have to write the formulas. Both libraries calculate indicators on price data.
| Tool | What it is | GitHub stars | Maintenance |
|---|---|---|---|
| TA-Lib Python | Python wrapper for the TA-Lib technical analysis library, with more than 150 indicators | 12.3k | Commits today |
| ta | Technical analysis library based on pandas and NumPy, implementing 43 indicators | 5.2k | No commits in over six months |
04 · Factors and performance: was the strategy any good
A finished backtest is the start. These libraries ask whether a factor predicted anything, and what the returns and drawdowns look like.
| Tool | What it is | GitHub stars | Maintenance |
|---|---|---|---|
| alphalens-reloaded | Library for analyzing how well stock factors predicted returns. Community-maintained fork of alphalens | 677 | Commits today |
| pyfolio-reloaded | Portfolio and risk analysis library. Community-maintained fork of pyfolio | 617 | Commits today |
| quantstats | Portfolio performance analysis library aimed at quant research | 7.8k | Last commit 30 days ago |
05 · Portfolio optimization: how to split the money
After you have chosen the names, decide how much goes to each.
| Tool | What it is | GitHub stars | Maintenance |
|---|---|---|---|
| PyPortfolioOpt | Portfolio optimization library, including mean-variance optimization, Black-Litterman, and hierarchical risk parity | 6.1k | Last commit 4 months ago |
| Riskfolio-Lib | Library for portfolio optimization and strategic asset allocation | 4.6k | Last commit 5 days ago |
| skfolio | Portfolio optimization library built on scikit-learn. Portfolio models can be cross-validated | 2.6k | Commits today |
How the three portfolio libraries differ is PyPortfolioOpt, Riskfolio-Lib, and skfolio. More are in the risk and portfolio category.
06 · AI agents: let a language model help with research
Several language-model roles split the analysis. They fit as research assistants. They should not place orders for you.
| Tool | What it is | GitHub stars | Maintenance |
|---|---|---|---|
| TradingAgents | Multi-agent LLM framework for financial trading. The project describes itself as research | 116.2k | Last commit 11 days ago |
| TradingAgents-astock | A-share multi-agent research framework adapted from TradingAgents, under Apache-2.0, for research and teaching | 3.8k | Last commit 2 days ago |
| Vibe-Trading | Personal trading agent with research and backtesting tools | 36.8k | Commits today |
Which of the three research agents fits which job is TradingAgents, TradingAgents-CN, and Vibe-Trading. More are in the LLM agents category. Market-app assistants, a general model, platform AI factors, and open-source research projects are sorted in how to choose AI stock-picking tools. That page is not a language version of this list.
07 · Trading bots are a different job
The tools above help you reach a conclusion. A trading bot sends orders. A mistake there is real money, so the choice is a different article. Freqtrade, OctoBot, Jesse, CloddsBot, and OKX's toolkit are compared in how to choose a trading bot. The same comparison in Chinese is how to choose a trading bot (Chinese). For an agent that talks to OKX, see OKX Agent Trade Kit: demo keys and read-only mode come before any live permission. More bots are in the trading bots category.
08 · Where to start
- US stocks: yfinance for data, plus one backtester. Store the daily update before you tune anything. See the price pipeline. To append EODHD bars onto an existing A-share table, see the EODHD walkthrough.
- China A-shares: AKShare for data, plus one backtester. Run one simple strategy, such as a moving-average cross, before you optimize. Account rules are on how to choose a quant trading platform. Data fees are on how to choose between Tushare and AKShare. Storing the daily bars is AKShare into SQLite. The same comparison in Chinese is Tushare and AKShare (Chinese). This page does not restate them.
- Once a strategy runs: add a performance library and look at the drawdown. That is more useful than writing another strategy.
- If you do not want to write code: most of the projects above need some Python. A charting site is a way to learn prices and indicators first. The walkthrough is the TradingView setup. The same article in Chinese is TradingView setup (Chinese). Disclosure: this site has an affiliate relationship with TradingView. A signup through our link may earn us a commission, at no extra cost to you. It does not change which projects are listed here.
The study order around this list is the beginner learning path. The full English directory, with maintenance status on every project, starts at all tools.
This is not investment advice. Features and licences follow each project's own docs. A backtest is not a forecast.