How to Install Msty

Msty is a desktop application that gives you a single interface for chatting with both local and cloud AI models. It bundles local model management (so you don't need to set up Ollama or llama.cpp separately) with support for connecting to online providers, split chats, and offline use. This guide covers installing it on Linux, macOS, Windows, and on a VPS for always-on access.

Before You Start

Msty ships as a desktop app for Windows, macOS, and Linux. Like any GUI application, it needs a display to run, so on a headless VPS you'll need a desktop environment and a remote desktop connection (VNC or RDP) first.

If you're only after local model inference and don't need the chat UI itself running remotely, you can also run a model backend headlessly and connect to it from Msty on your own machine, that avoids needing a graphical session on the server at all.

Installing Msty on Linux

1. Download the official Linux build (AppImage or .deb, depending on your distro) from Msty's website.
2. For the AppImage:

chmod +x Msty-*.AppImage
./Msty-*.AppImage

3. For the .deb package on Ubuntu/Debian:

sudo dpkg -i msty_*.deb
sudo apt-get install -f

4. Launch Msty from your applications menu, or run msty from a terminal.

Installing Msty on macOS

1. Download the macOS .dmg from Msty's website.
2. Open the .dmg and drag Msty into your Applications folder.
3. Launch it from Launchpad. On first open, macOS Gatekeeper may ask you to confirm you trust the app, since it's downloaded from outside the App Store.

Installing Msty on Windows

1. Download the Windows installer (.exe) from Msty's website.
2. Run the installer and follow the prompts.
3. Launch Msty from the Start menu once installation finishes.

First-Time Setup

Msty's setup wizard runs the first time you open it:

1. Choose local or online models. Msty can download and manage local models directly (similar to Ollama), or connect to cloud providers like OpenAI, Anthropic, or OpenRouter with an API key.
2. Download a local model (optional). If you want offline inference, pick a model from the built-in catalog; Msty handles the download and setup for you.
3. Add API keys (optional). For cloud providers, paste your API key into the provider settings panel.
4. Start chatting. Msty supports split-screen chats, so you can compare responses from two models side by side.

Running Msty on a VPS

Because Msty is a graphical app, running it persistently on a server means giving that server a desktop session:

1. Deploy a VPS and install a lightweight desktop environment:

sudo apt install xfce4 xfce4-goodies -y

2. Install a VNC server so you can connect to the graphical session remotely, then connect and install Msty as described above.
3. Size the plan for the models you intend to run locally: local inference is RAM- and CPU-heavy, so a larger plan pays off if you're loading anything beyond small models. If you're only using Msty as a client for cloud models, a smaller plan is plenty.

Frequently Asked Questions

Do I need a GPU to run Msty?

No, but local models run faster with one. Without a GPU, Msty falls back to CPU inference, which works but is noticeably slower for larger models.

Can Msty use models I already downloaded with Ollama?

Yes, Msty can connect to an existing local Ollama installation as one of its model sources, so you don't have to duplicate model downloads if you already have one set up.

Is Msty free to use?

Msty is free to download and use for local models. Cloud provider usage is billed by that provider based on your own API key and usage.

Get Started

To run local models without local hardware limits, deploy a VPS on CloudBlast sized for the model you want to run, connect over VNC, and install Msty in minutes. Check plans and pricing for the right RAM and CPU tier.