Msty
Desktop app for local LLMs — model management, RAG, and chat in one polished interface
📋 Overview
Msty is a desktop application for running local LLMs with a polished ChatGPT-like interface — featuring built-in model management, RAG with document upload, and multi-model chat. Unlike command-line tools (Ollama) or basic web UIs (WebUI), Msty provides a native desktop experience with drag-and-drop model installation, automatic GPU/CPU detection, and a clean conversation interface. The RAG feature lets you upload documents (PDF, TXT, markdown) and query them alongside normal chat. Available on Windows, macOS, and Linux with offline-first design.
✨ Key Features
- •Polished native desktop app — Windows, macOS, Linux
- •One-click model download with automatic quantization
- •Built-in RAG: document upload + vector search
- •Automatic GPU detection (CUDA, Metal, ROCm) with CPU fallback
- •Multiple concurrent conversations with history
- •Offline-first design — no cloud required
- •Drag-and-drop document ingestion for knowledge bases
🎯 The Problem It Solves
Local LLM setup remains fragmented — you need Ollama for model serving, a separate UI for chat, another tool for RAG, and manual config for each. Msty consolidates all of these into a single desktop application with a polished interface that non-technical users can navigate.
🔧 How It Works
Msty bundles Ollama model management with a native desktop chat interface. Models are downloadable with one click, with automatic quantization selection based on your hardware. The chat interface supports multiple concurrent conversations, conversation history, and model switching. RAG mode documents the sidebar, chunks them, embeds them into a local vector DB, and serves retrieval-augmented responses. GPU acceleration is auto-detected (CUDA, Metal, ROCm), with CPU fallback. No cloud required — everything runs locally.
🚀 Installation & Quick Start
Installation
Download from https://msty.appQuick Start
- Download and install
- Browse models in-app
- Download a model
- Start chatting
✅ Pros
- •Most polished local LLM desktop interface
- •One-click model management — no terminal needed
- •Built-in RAG with document upload
- •Automatic GPU/CPU detection and optimization
- •Offline-first — full data sovereignty
- •Cross-platform: Windows, macOS, Linux
- •Affordable Pro tier at /mo
❌ Cons
- •Smaller community than LMStudio or Jan
- •Advanced RAG features require Pro subscription
- •No API server mode — desktop-only
- •Very large models (70B+) still require high-end hardware
- •Proprietary license — no self-hosting or source access
💬 Practitioner Verdict
“Msty is the most polished local LLM desktop app available — the model management alone saves significant setup time, and the built-in RAG works well for personal knowledge bases. The trade-off: smaller communityLMStudio or Jan have larger user bases, advanced features require Pro subscription, and very large models (70B+) still need significant hardware. For non-technical users who want local AI without terminal commands, Msty is the best starting point.”
Self-Hosted (Free)
Open source, MIT/Apache licensed. Run it yourself.
⭐ Star & Clone on GitHubFree forever. Your infrastructure, your data.
Msty Pro
Unlimited models, advanced RAG
- Unlimited models
- Advanced RAG
- Priority support
📊 Specifications
- Language
- TypeScript
- License
- Proprietary
- Platform
- Linux, macOS, Windows
- Supported Models
- REST API, CLI
💰 Pricing Reality
Free tier: full chat, basic RAG, limited model downloads. Pro /mo: unlimited models, advanced RAG with reranking, priority support, early features. No API usage fees — you run your own models. Optional cloud backup/sync is extra.