
Desktop AI workspace that runs local and online models side by side. Free Studio plan and paid Aurum tier. For Windows, macOS and Linux.
Choose Msty if you want one window to test local and cloud models against each other, with prompts and documents kept on your own machine. It suits researchers and small teams more than developers who only need a server. Four product lines now exist, and the Stack knowledge layer is still unreleased.
Split chats that compare local and hosted models in one window on the free plan.
Switching cost is low because it sits on top of other engines, but personas, prompts and Knowledge Stacks held in the app add some lock-in.
Main trade-off
Many useful features are paid, and the product line is widening faster than the core app.
| Plan | Monthly price | What it includes |
|---|---|---|
| Studio Free | Free | Core Studio Desktop features: model chat, persona and crew conversations, Agent Mode, Knowledge Stacks and the prompt, persona, skill and media studios, some with limits |
| Aurum | $12.42/mo | Billed at $149 per user per year. Adds Msty Studio Web, Azure and Bedrock providers, Studio Assistants, Shadow Personas, Turnstiles, Insights and Forge Mode |
| Aurum Lifetime | Custom | One-time $349 per user payment for the Aurum feature set with lifetime updates |
| Enterprise and Teams | Custom | Custom pricing from 5 seats, with SSO, user and team management, role-based access control, audit log and shared team resources |
Studio is free to start. The Aurum tier is $149 per user per year and adds Studio Web, Azure and Bedrock providers and the advanced features. A lifetime licence is a single payment of $349 per user. Enterprise and Teams plans are custom and need at least 5 seats. Msty Go is free in beta. Checked on msty.ai 2026-10-07.
Msty Studio is a desktop app for Windows 10 and 11, macOS 11 and later, and Linux that connects three local engines (MLX, llama.cpp and Ollama) with hosted AI providers. Its split chats compare several models on one prompt. As of 7 October 2026, the vendor says SOC 2 and ISO 27001 certification are not yet complete.
Local and online models in one hub
Connects hosted providers with local engines named by the vendor as MLX, llama.cpp and Ollama, so one app covers both.
Split chats for model comparison
Sends one prompt to several models at once and shows the answers side by side, available on the free plan.
Knowledge Stacks
Retrieval over your own files and notes, with limits on the free plan and full access on Aurum.
Agent Mode and MCP tools
Lets a model call tools through the Model Context Protocol, with limited MCP use on the free plan.
Personas, prompts and media studios
Reusable personas, prompts and skills plus a media studio, so repeated work does not start from a blank chat.
Team controls on the Teams plan
SSO, role-based access control, an audit log and shared resources, sold from a 5-seat minimum.
Most users expect these and they are not currently supported.
| Tool | Price from | Free plan | Runs on | Rating |
|---|---|---|---|---|
| $12.42/mo | Yes | Web · Mac · Windows | -- | |
| $20/mo | Yes | Mac · Windows | -- | |
| $20/mo | Yes | Web · Mac · Windows | 4.0 | |
| Free | Yes | Mac · Windows | -- |
Msty Studio has a free plan. The Aurum tier is billed yearly at $149 per user and unlocks Studio Web, Azure and Bedrock providers plus the power-user features. Aurum Lifetime is a one-time $349 per user. Teams pricing is custom with a minimum of five seats, and Msty Go is free during its beta.
It covers desktop chat across both model types, split comparison, Agent Mode, and limited use of Knowledge Stacks, personas, prompts and MCP tools. Studio Web, Shadow Personas, Turnstiles, Insights and Forge Mode are not included. There is no stated time limit.
LM Studio is the closest desktop rival for browsing and running local models. Jan and GPT4All are simpler offline chat apps, and Open WebUI or AnythingLLM suit shared team workspaces and document chat. Choose Ollama or llama.cpp if you only need an engine.
They overlap on running local models. Msty adds split chats, personas, Knowledge Stacks and team plans around a mixed local and hosted model hub, while LM Studio is narrower and centred on discovering and running models. Try both on your own hardware, since speed depends on the machine.
Download the Studio installer for Windows 10 or 11, macOS 11 or later, or Linux from msty.ai. Add a hosted provider key or connect a local engine such as Ollama, MLX or llama.cpp, then start a chat and open a split chat to compare two models.
Msty is a desktop AI workspace that puts local and online models in one window. Its main product, Msty Studio, installs on Windows, on Macs with Apple silicon or Intel chips, and on Linux, and the vendor lists a free plan that covers chat with local and hosted models, Knowledge Stacks and Agent Mode. It sits in the same group as Ollama, LM Studio and Jan, but it is built around side-by-side comparison and team controls rather than a single model runner.
The core idea is a model hub. Studio connects to local inference engines, named on the vendor's pricing page as MLX, llama.cpp and Ollama, and to hosted providers, so you can send one prompt to several models in split chats and compare the answers. Around that sit persona and crew conversations, a prompt studio, a media studio, MCP tools and Knowledge Stacks, the vendor's name for its retrieval feature over your own files. If you already run models through llama.cpp or Ollama, Studio is a front end on top of them rather than a replacement.
The company now sells more than the desktop app. Msty Go runs scoped agents with Discord, Telegram and WhatsApp integrations and is in free beta, Msty Nexus is a model gateway for runtimes, keys and usage, and Msty Stack, a governed knowledge layer, is marked coming soon. For comparison, Open WebUI, AnythingLLM cover the browser-based team workspace and document chat, and GPT4All is the simpler offline option.
Who it suits: people who want one app to test local and cloud models against each other, researchers and analysts who work from their own documents, and small teams that need shared prompts and access control. Hosted assistants such as ChatGPT and Claude are the usual yardstick for the local models you test, and Hugging Face is where many of those open models are published. Developers who only want an API endpoint will find llama.cpp lighter. Our guide to local coding models covers which models are worth loading once the app is installed.
Pricing has a free Studio plan, a paid yearly Aurum plan that adds Studio Web and the advanced features, a one-time lifetime licence, and custom Teams plans from five seats. Regulated teams should ask the vendor for its current controls in writing before buying, because certification is still in progress.
Product facts on this page (pricing, plans, integrations, compliance) are read from Msty's own pages and documentation. The verdict is Hokai's reading of them, not the vendor's claim.