MCP server for bean-query

[Tammo van Lessen]

Hi all,

In the LLM world, everyone seems building MCP servers right now. These enable LLMs (such as Claude) to connect directly to external tools or APIs.

I wrote an experimental MCP server for beanquery over the weekend as a demo and Claude seems to be able to do BQL quite well.

https://github.com/vanto/beanquery-mcp

Screenshot from a simple conversation: https://github.com/vanto/beanquery-mcp/blob/main/docs/screenshot-expanded.png

This is certainly only a PoC, and you should be careful which data you want to share with which LLM.

Maybe someone here will find it interesting.

Best,
  Tammo

[Martin Blais]

We live in wild times

I love it

I've added an "AI" section to the contrib document with a link to your project

(related)

I've had an idea recently: 

- take your document cache

- extract (account or directory, date, text-conversion-of-pdf document)

- use a pretrained embedding model to train a classifier from (text -> date) and (text, directory or account)

- automatically move a new list of pdfs from ~/Downloads/ to the directory hierarchy

Basically "store these documents where other ones like it are located"

My next session of vibecoding awaits...

--
You received this message because you are subscribed to the Google Groups "Beancount" group.
To unsubscribe from this group and stop receiving emails from it, send an email to beancount+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/beancount/b0af3640-b79a-4506-969c-40742a79c869n%40googlegroups.com.