MindMap Chat

Set up LM Studio

MindMap Chat can talk to LM Studio, a local AI server with a desktop UI of its own. Run it on your own machine, or on one capable PC that serves every device on your network. Either way, your map content never has to touch the public internet — and because the AI runs locally, there are no per-token fees or subscriptions.

Prefer a runtime with no UI of its own, or need automatic setup on Linux? Use Ollama instead — the app supports both, and compares them here.

On Windows and macOS the app can install and start LM Studio for you — press Save in Settings → AI and it installs LM Studio, starts its server, and downloads your chosen model. On Linux LM Studio ships only as an AppImage, which cannot be installed unattended, so that one is a short manual setup.

LM Studio installs its command-line tool on first launch. If the app tells you the lms tool is not set up yet, open the LM Studio desktop app once and press Retry. Nothing else is needed.

Install LM Studio yourself

Windows

winget install -e --id ElementLabs.LMStudio

Or download the installer from lmstudio.ai/download.

macOS

brew install --cask lm-studio

LM Studio for macOS requires Apple Silicon. On an Intel Mac, use Ollama instead.

Linux

Download the AppImage from lmstudio.ai/download, make it executable, and run it once. It needs FUSE — on Debian/Ubuntu, sudo apt install libfuse2. 

chmod +x LM-Studio-*.AppImage
./LM-Studio-*.AppImage

Launch it once

Open the LM Studio app. Besides being where you browse models, the first launch is what installs its lms command-line tool, which MindMap Chat uses to start the server and download models.

Download a model

From a terminal, or from LM Studio's own model browser:

lms get qwen/qwen2.5-7b -y
# or a smaller, lighter option:
lms get google/gemma-3n-e4b -y

Start the server

lms server start -p 1234

Point the app at it

Open Settings → AI, choose LM Studio as the provider, keep the Base URL as http://localhost:1234/v1, pick the matching Model, and press Save. The model's badge should read installed.

The first reply from a model can take a while: LM Studio loads it into memory on demand. Later replies are fast.

One server for your whole network

You don't need LM Studio on every device. Run it once on a single capable PC — plenty of RAM, and ideally a GPU — and let every other laptop, tablet, and phone on the same network use it.

On the host machine

  1. Start the server bound to every interface. By default it listens only on localhost, which other machines can't reach: 
    lms server start --bind 0.0.0.0 -p 1234
  2. Download the models you want to share (see above).
  3. Allow inbound TCP port 1234 through the machine's firewall — the user guide has the exact command for Windows, Linux and macOS, and what to watch out for.
  4. Note the host's LAN address, for example 192.168.1.50.

On every client device

  1. Open Settings → AI and choose LM Studio.
  2. Set the Base URL to http://192.168.1.50:1234/v1 (your host's address).
  3. Pick a model the host has downloaded, and press Save.

Clients don't install anything — they just point at the shared host. Because LM Studio downloads models through its own command-line tool, the app can only fetch a model on the machine running LM Studio; for a shared host, download the model there. The app says so instead of offering a download that could not work.

Keep your endpoint on your own network

Where LM Studio runs decides where your data goes. Everything the AI sees — your map outline, notes, and chat — is sent to whatever Base URL you configure. MindMap Chat tells you which of three situations you're in:

⚠ Don't point the app at an LM Studio endpoint on the public internet unless you run and trust that server.

Doing so streams your private map content off your network, where it may be logged or retained. Keep the server you point at on your own machine or trusted local network. Your local database is always encrypted at rest, but that protection does not extend to content you send to an off-network server — which is exactly why the server you choose matters.

Troubleshooting