Musical AI
YOUR SPACE FOR AI

Your models.
Your rhythm.

From the first thought to the finished result.
Local models, a team of agents and your tools — together in one app.

Download Musical AI View on GitHub ↗

Windows · macOS · Linux · No subscription required for local inference

MUSICAL AI / WORKSPACELess noise. More room to think.
Musical AI desktop interface with projects and a chat composer
Local when you want independence.Cloud when you need larger models.

01 / POSSIBILITIES

Build your own way to work.

Choose the models and tools that fit your task.

01

Models on your computer

Run local models through Ollama on every supported platform. Windows also offers a built-in engine and a one-step setup.

Local inference works offline after downloading models
02

One request. Several agents.

A coordinator assigns suitable parts of a task to specialists. Watch their progress while the main model combines their results.

Up to two specialists · Speed depends on your model and hardware
03

Tools within reach

Connect MCP servers and Markdown Skills. The built-in Workspace lets models read your projects with your chosen permissions.

MCP · Skills · Action approvals

02 / INSIDE THE APP

One workspace. More possibilities.

Start with local chat. Add tools and cloud models whenever you need them.

Musical AIFEATURE OVERVIEW
ON YOUR COMPUTER

Start with one model.

Windows Quick setup downloads the engine and Qwen 1.5B. On macOS and Linux, install Ollama and choose a model.

Qwen 2.5 · 1.5BLocal
OllamaWindows, macOS, Linux

Models download separately. Larger models need more memory.

03 / YOU ARE IN CONTROL

Know what is happening.
Choose what happens next.

Local or cloud

Local requests run on your computer. With Cloud selected, messages and attached context go to Ollama, as indicated in the app.

Approve or automate

Choose tool permissions. Planning mode disables MCP calls; approval mode shows each proposed action before it runs.

Make it yours

A calm interface, animated strings and Discord Activity. Shape your workspace around the way you work.

READY FOR YOUR FIRST IDEA?

Your next step starts here.

Choose your system and install Musical AI.

Version 1.19.0

On macOS and Linux, install Ollama for local models or connect Ollama Cloud. Windows also supports the built-in engine.
macOS builds are unsigned and not notarized by Apple.

View on GitHub ↗All releases ↗

A FEW MORE QUESTIONS

Before you start.

Do I need a subscription?

Local inference requires no subscription. Ollama Cloud is a separate service with its own free limits, plans and payments on ollama.com.

Do I need a powerful GPU?

Small models can run on a CPU. Speed and model size depend on your hardware and available memory. Cloud models do not need a local GPU.

What does Quick setup install?

On Windows x64, it downloads llama.cpp and Qwen 2.5 1.5B: about 1.14 GB, plus Microsoft Visual C++ components if needed. macOS and Linux use Ollama for local models.

How do I install on Linux or Mac?

Linux: install the .deb on Ubuntu/Debian, or make the AppImage executable and run it (FUSE may be required). Mac: open the DMG and drag Musical AI to Applications. macOS may require approval in Privacy & Security for the unsigned app.

Where can I report a bug?

Open GitHub Issues and include the app version, model and steps to reproduce. GitHub Issues ↗