Another way to use local llm, have an MCP server that talk to a Qemu computer. What do you think?

I think is nice to contain the MCP into a Qemu enviroment where the LLM can do whatever ... here is doing GDB on a LVGL program.
https://github.com/leonardosalvatore/llama.cpp.debugger

[-]

You dont need an mcp to instruct your llm to use virsh remotely/locally

[-]

yes if I need to call GRPC, MQTT methods that are not part of this open source project.

[-]

Also I found useful that the tools "description" will give a context to the llm.
But I'm still learning =]

[-]

LLM probably knows what each parameter do. Even it is not it can look at help and man pages

[-]

Yes. For standard tool you are right. Not for a custom API.

[-]

Again virsh is not a custom api. You dont need mcp for it.

[-]

I know what is for but doesn't work when you need to call custom API.

[-]

I dont need to call a custom api for a qemu which your post is for.

[-]

exactly you are right.
you don't need to control a linux os.

But yes for GRPC and MQTT which are not part of this post but are part of project i'm working on.

But i already mention in my first reply.

[-]

No need to call api anything shit you cant just use virsh cli and any llm can figure it out without any mcp

[-]

I agree. And I find that MCP is unnecessary for more than half the uses I encounter