Use any AI you like
Plug in the big names like OpenAI, Anthropic, and DeepSeek, connect tools you already run such as Ollama or LM Studio, or use an AI built right into the app. Six easy ways to connect.
MAOS Mac OSmacOS build
MultiAgentOS for Mac is a friendly desktop assistant you can chat with all day. Ask it to draft an email, dig through your files, research a topic, or tidy up a folder, and watch it do the work right there on screen. It runs on any Mac, keeps your conversations private, and even works with no internet at all.
What you can do with it
Chat with it like a colleague, hand it a task, and let it open the right tools, work through your files, and take action on your Mac, always with your say-so.
Plug in the big names like OpenAI, Anthropic, and DeepSeek, connect tools you already run such as Ollama or LM Studio, or use an AI built right into the app. Six easy ways to connect.
An AI comes built in, so you can start chatting straight away with no account and no setup. On a plane or off the grid, it keeps working, and your prompts and files never leave your Mac.
Any passwords or sign-in keys are stored safely in the Mac's built-in Keychain, never in plain text. Nothing is tracked, no logs are kept, and your chats stay on your machine.
It can move the mouse, click, type, scroll, take screenshots, and open apps to carry out a task end to end. You stay in charge, since it asks before every action it takes.
Need a web browser, a code editor, a file viewer, or a document open? The assistant slides the right panel in beside your chat when a task calls for it, then tucks it away again.
Got a larger task? The assistant can spin up little helpers to tackle parts of it in the background, each with clear limits you set, then bring the finished results back to you.
A closer look
Set up your AI, open files to work on, and keep typing to the assistant, all in the same place. No hopping between half a dozen apps to get something done.
Connect a service like OpenAI or Anthropic, use an AI on your own machine, or go fully offline, all from one simple settings panel.
Open your files and folders right next to the chat, so you can tweak something and keep talking to the assistant without losing your place.
How to use MAOS Mac OS
A good first session is simple: install the app, decide which Mac permissions you want to grant, connect an AI, then ask for one visible task while you watch.
Move the app to Applications if you want it alongside your other Mac apps, then open it and pin it to the Dock if you plan to use it daily.
Screen Recording lets the assistant understand what is visible. Accessibility lets it click and type. File access lets it work with folders. You can leave any permission off until a task actually needs it.
Use a hosted model when you want maximum reasoning, or connect Ollama, LM Studio, llama.cpp, or a GGUF file when the work should stay private on your Mac.
Try: "Open my Downloads folder, list duplicate-looking files, and stop before deleting anything." This lets you see how the assistant reads, reasons, and pauses before sensitive actions.
Attach a file, enable an MCP server, or open the browser sidecar only when a task needs it. A smaller tool set makes the assistant easier to supervise.
When a workflow works, save the wording: the folder, the rules, what to summarize, and where the assistant must stop for your approval.
Beginner Mac workflows
Treat the assistant like a careful coworker on your Mac. Give it one clear job, one source of context, and one stopping rule, then expand only after you trust the workflow.
Attach a folder of notes, PDFs, or meeting files and ask: "Summarize the key decisions, list open questions, and do not edit the files." Use a local model when the material should stay on the Mac.
Open Settings, choose Local Server or Local AI, point at Ollama, LM Studio, llama.cpp, or a GGUF file, then send a short test prompt before adding files, screenshots, or desktop control.
Open a page in the browser sidecar and ask for a comparison, summary, or source list. Tell the assistant to quote the page title, keep links, and stop before logging in or submitting forms.
Ask it to inspect Downloads, Desktop, or a project folder and create a cleanup plan. Keep the first pass read-only, then approve moving or renaming files one batch at a time.
For command-line work, ask for the command, the reason, and the expected output before anything runs. This is useful for Git checks, file searches, dependency installs, and local server setup.
When a prompt works, save it as a repeatable checklist: folder to read, model route, allowed tools, output format, and the exact actions where the assistant must pause for approval.
Mac questions
No. You can connect cloud AI providers when you want their reasoning, but MAOS Mac OS also supports local servers, local GGUF models, and a bundled offline runtime for private work.
Start with the smallest set. Use file access for document work, Screen Recording for visible desktop context, and Accessibility only when you want the assistant to click, type, or move through apps.
Yes. You can connect Ollama, LM Studio, compatible local servers, CLI pipes, terminal commands, and local GGUF model files from the model settings.
Use a read-only task first, such as summarizing a folder, drafting a reply, researching a page, or listing cleanup suggestions while telling the assistant to stop before deleting, sending, or changing anything.
Built for the Mac
One download works on every modern Mac, new or old, and fits in naturally with the menu bar, the Dock, and the privacy features you already trust.
Support the project
Chip in whatever you like to support the project and download the app straight away. Each download link works once, just for you.
Get the software
Donate whatever feels right, pick the app and your computer, and download it right after checkout. Each download link works once and only for a short while.