CNEquity — Claude Code MCP server: China A-share data locally
Gives an AI agent prices, fundamentals, flows and events for China A-shares from a local Parquet lake. Useful for repeated backtests and long-term history.
claude mcp add cnequity -- uvx cnequityWhat CNEquity does
CNEquity pulls equities, futures, financial statements, corporate events and capital flows into a local Parquet lake. The agent then reads it over MCP in read-only mode. Inside are incremental collection, resume after failure, quality audits and one query layer. It registers 55 datasets: 47 curated and 8 derived, grouped into levels L0–L9.
You need Python 3.10+ on macOS, Linux or Windows. No account or token. A one-minute demo profile fetches 5 stocks over the last 30 trading days into a separate directory, while a full-market build over 3 years takes hours and survives a restart. Minute bars, ticks and derivatives are off by default; enable them and check the source's coverage.
What people come here for is control over口径: raw prices and adjustment factors are stored separately, delisted names stay in the historical universe, and statement queries distinguish strict PIT from after-the-fact reconstruction. If you backtest on today's list of tickers, you drop the delisted ones — that is a reason to check your pipeline. Data sits in open files, readable through Python, DuckDB, Polars, the console and MCP.
What it helps with
- build a local A-share price lake with no account or token
- resume a download after a network drop without redoing finished work
- keep raw prices and adjustment factors apart in one query
- retain delisted names in the historical universe
- check which statements were visible on a rebalance date
Example prompts for the agent
- Connect cnequity and tell me which day 宁德时代 had its highest closing price over the last 30 trading days
- Using cnequity, show which names gained the most over the past month from daily bars
- Run cne init and cne check, then tell me which steps did not finish
How to install CNEquity
Platform CLI
claude mcp add cnequity -- uvx cnequityuvx runs the command named after the package; if the README names another one, use uvx --from cnequity <command>
Platform CLI
codex mcp add cnequity -- uvx cnequityuvx runs the command named after the package; if the README names another one, use uvx --from cnequity <command>
Config file
{
"mcpServers": {
"cnequity": {
"command": "uvx",
"args": [
"cnequity"
]
}
}
}uvx runs the command named after the package; if the README names another one, use uvx --from cnequity <command>
Config file
{
"servers": {
"cnequity": {
"type": "stdio",
"command": "uvx",
"args": [
"cnequity"
]
}
}
}uvx runs the command named after the package; if the README names another one, use uvx --from cnequity <command>
Platform CLI
gemini mcp add cnequity uvx cnequityuvx runs the command named after the package; if the README names another one, use uvx --from cnequity <command>
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about CNEquity
How do I install CNEquity in Claude?
Run claude mcp add cnequity -- uvx cnequity. The server comes up as a read-only MCP server.
What does CNEquity do?
It collects prices, fundamentals, flows and events for China A-shares into a local Parquet lake and serves them to an agent over MCP. It also resumes downloads after a failure and audits data quality.
What do I need to run CNEquity?
Python 3.10+ on macOS, Linux or Windows. No account or token. Minute bars, ticks and derivatives are off by default.
