[DISCUSSION] Superpower coding agents with C# LDM MCP server

[FrancisHogle]

I have been experimenting with various generations of AI-assisted coding. One notable gap in multiple models is up-to-date knowledge of the status C# language design. Manually reading all the discussion, proposal and meeting notes is no small task, and seems well suited to the agent vision.

At present, users cannot reliably have meaningful agent interactions of the form "why does this work this way?" / "how did this get missed?" / "which approach might the community favor...?"

If the information was available before the model's original training date, then it can be a "friend who has read every meeting note." However, models have gaps on the most current discussions and histories (sometimes up to two years of missing context).

Any of us could webscrape github and msdn, but that's diffuse, redundant and error prone. Should Microsoft provide this type of resource - an MCP server that covers the entire corpus of C# LDM knowledge and history?

This would greatly improve agent relevance and accuracy. It would also provide a more reliable "search for similar discussions" :)

One can also imagine standing up MCP servers for other aspects of .Net, Azure, Windows etc, enabling expert assistance via agent discussions.