VS code: https://code.visualstudio.com/docs/intelligentapps/models

> Models downloaded locally from repositories like Ollama and ONNX.

It's absolutely possible, but on an M4 mac the largest / slowest model I could feasibly run was very inferior, compared to the default paid Cursor experience.

I tested with Kilo Code: https://kilocode.ai/ -- it's a VS Code / Cursor extension.

To host models on desktop, there's this: https://ollama.com/

I've been looking for this too. It seems to me as though all the ide's are trying to sell the llms as a service or trying to lock you in by downloading llms through their ide. I have been downloading llm's from huggingface as gguf files and would like to use those downloads (and running them through https://github.com/oobabooga/text-generation-webui). It is possible to run those llms as a local api using something like llama-cpp-python (https://pypi.org/project/llama-cpp-python/) and would prefer to use something like that method. Zed (https://zed.dev/), which is now available on windows might be able to do it, but i'd rather use something (foss) that doesn't have a pricing model (the development focus will always be upon those who pay). tbh i'm getting a bit sick of changing ide's, as their support changes, and really would prefer not to use (microsoft) visual studio code which seems to be cornering the market. Starting to think i'm going to try to learn emacs, with https://github.com/karthink/gptel looking as if it would meet my needs.

try zed

I don't know about out-of-the-box but there's several popular extensions for both Jetbrains and VS Code that let you use local models.

Roocode

https://github.com/RooCodeInc/Roo-Code

Continue

https://github.com/continuedev/continue

You could try Zed IDE with Ollama