Local AI: pick a runtime
New in v1.18.1: Three more local models to pick from — Muse Glimmer, Qwen 3.8 (27B) and, on LM Studio, Bonsai 27B. — all changes
MindMap Chat never sends your maps to a cloud AI service. The model runs on your own hardware, through one of two local runtimes — Ollama or LM Studio. Both give you the same features and the same curated models; pick whichever you already have, or whichever setup you prefer.
| Ollama | LM Studio | |
|---|---|---|
| Set up for you | Windows, macOS, Linux | Windows, macOS — on Linux it is a short manual setup |
| Fetches models for you | Yes, including to a server elsewhere on your network | Yes, on the machine running it |
| Has a desktop UI of its own | No — it is a background service | Yes, with a model browser and chat |
| Default address | http://localhost:11434/v1 |
http://localhost:1234/v1 |
| Guide | Set up Ollama | Set up LM Studio |
Not sure?
Choose Ollama if you want the app to handle everything, on any desktop OS, with nothing else to learn. Choose LM Studio if you already use it, or if you want a graphical model browser and a place to try models outside the app.
You can switch at any time in Settings → AI: pick the other provider and the app re-points itself at that runtime's address and shows its own model list.
On a phone or tablet
Neither runtime runs on Android or iOS. The app shows just the server address, so run Ollama or LM Studio on a desktop on the same Wi-Fi and point your phone at it — both guides cover sharing one server across your network.
Whichever you pick, the map content in a prompt goes only to that server. See the privacy page for exactly what is sent where.