Msty

Msty is a desktop chat client for local and remote LLMs. Where LM Studio bundles its own inference engine and model browser, Msty is deliberately a thin client: it talks to any OpenAI-compatible endpoint — Ollama, Mullama, LM Studio, llama.cpp server, even OpenAI and Anthropic — and gives you a clean, fast chat window on top. The install is a single ~80MB download, startup is near-instant, and the UI is the kind of native-feeling thing that respects your screen real estate instead of imitating a web app. For someone who already has an inference backend running and just wants a nice front-end — or who wants to flip between local and cloud models in one window — Msty is the lightest, least-opinionated option in 2026.

Key Features

• Backend-agnostic by design. Add any number of OpenAI-compatible endpoints and pick between them per chat. Local Ollama, a remote Mullama daemon, and OpenAI’s API can coexist in the same model dropdown. No vendor lock-in, no reconfiguration when you swap engines.
• Small, native, fast. The binary is a fraction of LM Studio’s size and opens in under a second. There is no Electron footprint and no bundled Python runtime; it is a desktop app, not a wrapped browser.
• Local RAG without a server. Point Msty at a folder of PDFs, Markdown, or text files and it indexes them in-process for retrieval-augmented chat. It is not a full document workspace like AnythingLLM, but for “chat with these five PDFs” it removes every setup step.
• Multi-model conversations and branching. Fork a conversation, switch models mid-thread, and compare answers side by side. Useful for evaluating local vs. cloud or two quant levels against each other.

When to Use Msty

Use Msty when you already run an inference server (Ollama, Mullama, llama.cpp) and want a polished desktop client instead of a browser tab. It is also ideal for hybrid setups where you mix local models with a paid API key and want one window for both. Choose LM Studio if you want the engine and model browser bundled together; choose Open WebUI if you need a server-side UI for multiple users.

Setup

# Download the installer for your platform from https://msty.app
# Windows: MstySetup.exe  (scoop: scoop install msty)
# macOS:   Msty.dmg       (brew: brew install --cask msty)
# Linux:   Msty.AppImage or .deb

# Then point it at a running backend:
#   Ollama:    http://localhost:11434
#   Mullama:   http://localhost:8080/v1
#   llama.cpp: http://localhost:8080/v1

No separate Python environment or Docker container required. The first launch walks through adding a provider; for local RAG, drag a folder onto the sidebar.

See also

• Ollama — the most common backend Msty connects to
• Open WebUI — server-side multi-user alternative
• All chat & UI tools — full catalog of frontends