Local LLM models

[ciroiriarte]

Hi!, is it possible to offload some tasks to local LLM services?

[itxashancode]

I ran into this same thing last year when I was trying to keep sensitive data local. The short answer is yes - you can absolutely hook up a local LLM like Ollama or LM Studio to Claudecode. You'd need to modify the model config to point to your local endpoint instead of the cloud API. Check the config.json or wherever the model URLs are defined - swap out the API base URL for something like http://localhost:11434 if you're using Ollama. The trick is making sure the request format matches what your local server expects. I'd start by testing with a simple curl to your local endpoint first, then wire it into the app. Not certain if Claudecode has a built-in toggle for this, but the config is usually the place to look.

[lloydsilvertwo]

I'd love to see this as well. It can obviously be done outside of OMC but just naturally fits if this were added.