Bring your own model
Six connection modes: API Key, Local Server, CLI Pipe, OAuth, Terminal, and Local AI — OpenAI, Anthropic, DeepSeek, Ollama, LM Studio, llama.cpp, and on-device GGUF files.
MultiAgentOSfor macOS
The macOS edition of MultiAgentOS runs on the shared Avalonia / .NET shell — Apple Silicon and Intel. Bring any LLM, keep your keys in the Keychain, let supervised agents drive your Mac, and run entirely offline with a bundled llama.cpp runtime.
What you get
One native shell that talks to any model, opens sidecars on demand, and acts on your machine under supervision.
Six connection modes: API Key, Local Server, CLI Pipe, OAuth, Terminal, and Local AI — OpenAI, Anthropic, DeepSeek, Ollama, LM Studio, llama.cpp, and on-device GGUF files.
A zero-setup bundled llama.cpp runtime means MultiAgentOS works with no internet and no API key — prompts, files, and history stay on your Mac.
API keys and OAuth tokens live in the macOS Keychain, never in plaintext on disk. No telemetry, no log files — conversation history is in memory only.
CGEvent / CoreGraphics desktop control lets the agent move, click, type, scroll, screenshot, and launch apps — every action gated by your approval.
Browser, desktop, code editor, terminal, file, and document sidecars open from the same frame when a task needs them, then close again.
Delegate bounded work with a goal, a tool allow-list, recursion limits, and a turn budget — then hand the result back to the main thread.
Interface details
Configure connections, open code sidecars, attach files, and keep the prompt controls visible without jumping between separate apps.
Set up API keys, local servers, CLI pipes, OAuth providers, Terminal templates, and local AI from one sidecar.
Edit files and folders beside the conversation, with model routing, permissions, and tools still visible.
Built for the Mac
Distributed as a universal app that detects Apple Silicon and Intel automatically, with the system integrations Mac users expect.
Supporter thank-you
Donate $10 toward ongoing development and choose the macOS edition of MAOS as your thank-you download.