September 1, 2026 at 10:50 am
Hi guys and gals,
I've created a system that automatically allows you to "generate" MCP tools based upon SQL Server CRUD and automatically generated endpoints. The thing is based upon meta programming and a DSL, which I had to create my own LLM for (yes, really!)
The point is you connect it to your database, the system reads meta data, and generates HTTP endpoints automagically. These endpoints are automatically added to the MCP server, allowing you to generate everything required to search, read, create, delete, and update rows from your database.
In addition it contains "SQL to MCP tool" components, and it can also convert an OpenAPI specification into tools, etc ...
I'd love feedback from the community. You can find it here.
September 2, 2026 at 11:10 am
Thanks for posting your issue and hopefully someone will answer soon.
This is an automated bump to increase visibility of your question.
September 2, 2026 at 11:36 am
This is the same system from 5 years ago with some additions. Are you looking for a critical evaluation of your design tradeoffs? "full security" seems overly optimistic imo
Aus dem Paradies, das Cantor uns geschaffen, soll uns niemand vertreiben können
September 2, 2026 at 12:00 pm
"The same as from 5 years ago" is a bit exaggerated. 6,500 more commits for instance the last 5 years?
The security model is based upon RBAC, in addition to static analysis of code. For instance, the integrated LLM cannot, not even in theory hallucinate (function invocations). Which literally implies that the AI will never return "semantically wrong code", and this is not even possible in theory - Since all code returned from the LLM is security checked statically before being returned ...
Viewing 4 posts - 1 through 4 (of 4 total)
You must be logged in to reply to this topic. Login to reply