CNEquity MCP Server: Setup for Claude, Cursor and Other Apps

• Shows what data your lake holds and how to read it

  Lists the datasets in your local lake, how far back each one goes, and the rules that make answers correct, such as price adjustment and point-in-time data. The project says to call it first in every session.

  Try: “What data do I have in the lake, and how far back does it go?”

  Files · describe_lake

• Turns a stock name into its code

  Looks up a company name, such as Moutai, and returns its stock code, such as 600519.SH. Delisted stocks are included and flagged with a delist date.

  Try: “What is the stock code for CATL?”

  Files · resolve_symbol

• Reads daily, index and minute price bars

  Returns price bars for stocks and indexes. You can pick price adjustment and the stock universe, and it warns you when you ask without choosing an adjustment.

  Try: “Which day had CATL's highest close in the last 30 trading days?”

  Files · query_bars

• Reads financial statements as they were known on a date

  Returns financial statement items as of a date you must give, so a historical question is not answered with today's restated numbers. Rows added by a later backfill are marked as not exact.

  Try: “Show Kweichow Moutai's net profit as it was known on 2024-03-31.”

  Files · query_fundamentals

• Reads any other dataset by date and symbol

  Reads any other registered dataset, such as fund flows, margin balances, dragon-tiger lists or share unlock schedules, filtered by date window and stock symbols, a page at a time.

  Try: “Show northbound fund flows for the last 10 trading days.”

  Files · query_dataset

• Runs one read-only SQL query across the whole lake

  Runs a single DuckDB SELECT over all datasets for rankings, averages and joins, so the assistant does not page rows through the chat. It rejects anything that is not a single SELECT and limits file access to the lake folders.

  Try: “Rank the 10 stocks with the largest 5-day gain in the last month.”

  Files · run_sql

Watch out for

• HTTP mode with a tunnel exposes your data

  For ChatGPT you run an HTTP mode and a public tunnel, and the token sits in the URL. Anyone with that URL can run read-only SQL on your lake, so keep it private and stop the tunnel when done. github.com

• Vendor news text sits in the lake

  The lake holds news_headlines and flash_news_wire, which are third-party text the assistant reads. The project says SQL can be influenced by content ingested into the lake. run_sql accepts only a single SELECT and is limited to the lake folders, but the project says this is not an operating-system sandbox. github.com

• Live mode fetches data on demand

  With the live option, lookups and daily-bar requests are fetched from data sources on the spot without writing to disk. Answers there are unadjusted, so use the full lake for price history. github.com

• Data comes from third-party sources

  The lake is built from public market data sources, and the project lists source limits and history horizons in its dataset notes. Check coverage before relying on it. github.com

Security advisories