Happy slice of mcp pizzaMCP.pizza

xiyan_mcp_server

MCP.Pizza Chef: XGenerationLab

Point it at a MySQL or PostgreSQL database, ask what last quarter's sales looked like, and it works out the query, runs it, and hands back a small table trimmed to about a hundred rows. Asking is the only thing on offer: there is no option for adding or editing records. The default arrangement sends your table and column names to a hosted model and needs a paid key, while a local-model mode keeps everything in-house. Setup means writing a settings file.

Data
Spreadsheets

Use This MCP server To

Ask how many orders came in last month Get a sales summary without writing a query Explore a database I have never seen before Check which customers spent the most this quarter Peek at sample rows from an unfamiliar table

README

xiyan_mcp_server FAQ

Is this still maintained?
The last commit was February 2026, so it has been quiet for several months. It is not archived and is still published on PyPI.
Do I need a paid key?
In the default setup, yes, because a hosted model writes the query. A local-model mode removes the key but asks you to run the model on your own hardware.
Which databases does it work with?
MySQL and PostgreSQL. Nothing else is supported.
Can I use this to get a sales summary without writing a query?
Yes, that is its whole purpose. You ask in ordinary language and it composes and runs the query behind the scenes.
Can it add or change records?
It offers nothing for that; the single option it exposes is for answering questions. No hard block exists, though, so connect it using an account that can only read.
Is my data sent anywhere?
In the default mode your table and column names go to the hosted model that writes the query. The local mode avoids that completely.
How hard is setup?
More than pasting a snippet. You install it with Python 3.11 or newer and write a separate settings file holding your model and database details.
Which apps does it work in?
Claude Desktop, Cursor, Cline, Goose and Witsy are all documented.
How much data comes back?
Results are trimmed to roughly a hundred rows, so it suits summaries and spot checks rather than bulk exports.