Setup
Most of the tools in the store talk to a model running locally through LM Studio. It is free, it runs on your hardware, and setup takes about five minutes. Nothing you do leaves your computer.
Download it for macOS, Windows, or Linux and open it once. It is a normal desktop app.
In LM Studio, open the search tab and pull one model to start. If you are not sure which, a small Qwen model is a safe first choice: quick to download and it runs comfortably on most recent machines. You can always add more later.
Some tools need a vision model (one that can see images). The per-tool list below says which.
Open the Developer tab (or the local server panel) and start the server. LM Studio then serves an OpenAI-compatible API at:
http://localhost:1234
That address is what every tool in the store looks for. Leave LM Studio running while you use them.
Load a model in LM Studio, then check the server is answering. In a terminal:
curl http://localhost:1234/v1/models
If you get back a list of models as JSON, you are set. Open any tool in the store and it will connect.
Which model for which tool
Any chat model. Add a vision model too if you want to drop images into a node.
A vision model, for example qwen/qwen3-vl-4b. It reads the image tiles.
Any chat model that is decent with code. A mid-size Qwen works well.
Any chat model. The phone connects to the same LM Studio server on your Mac.