hello245m/free-stockdb

free-stockdb is a local data engine for Chinese A-share stock market data, covering daily, weekly, and monthly candles as well as minute level and tick level prices for stocks and ETFs. Instead of calling a remote server every time you need a price or an indicator, it downloads and stores market data on your own computer, so research and backtesting run against a local dataset rather than a live network connection. The project argues that the real bottleneck in large scale backtesting is not writing trading strategies, it is the data engineering behind them: downloading years of data for thousands of stocks, cleaning it up, adjusting prices for stock splits and dividends, and organizing it so it can be queried quickly. free-stockdb handles this ahead of time. After an initial full download, later updates only fetch data that has changed, using compressed transfers that can resume if interrupted. Once the data is synced, later queries, calculations, and backtests all run against the local copy, so nothing depends on the speed or availability of the original data source. The dataset comes with 39 built in technical indicators, such as moving averages, MACD, and KDJ, plus five ways of calculating a custom index, computed with a C++ engine the project says runs several times faster than a typical Python approach. You can reach the data five different ways: a Python library for research and backtesting, an HTTP API for connecting from any programming language, an Excel or WPS macro, a plain HTML page for browsing without writing code, and an MCP connector that lets AI tools like Claude, Cursor, or Windsurf query the local data directly. The data source is treated as replaceable. You can point the sync configuration at your own provider, an internal server, or a shared folder, and once data is downloaded it keeps working offline even if the original source disappears. Compressed storage keeps disk usage well below plain CSV or database files, with a full setup needing roughly 5 to 20 gigabytes depending on whether minute level data is included. Prebuilt releases are available for Windows, macOS, Alpine Linux, and manylinux, while building the C++ components from source needs CMake, a C++17 compiler, libcurl, and OpenSSL.