Streaming replies
Answers appear as they are written. Turn on thinking to see the model’s reasoning separately.
Your private AI
AiDesk Helper is a free chat app for the AI models you run with Ollama. Your questions, your files and your answers never leave your computer. No account, no subscription, no API key.
Requires Ollama (free) · macOS on Apple Silicon

Features
Except the part where your words go to someone else’s server.
Answers appear as they are written. Turn on thinking to see the model’s reasoning separately.
Pick a model under the chat title. Models already loaded in memory are listed first.
Settings ▸ Models shows what you have, installs new ones with live progress and removes the ones you no longer need.
Paste or drop up to 4 images for vision models, or generate pictures with an image model.
In the desktop app, AiDesk can read, search and change files in the folder you open.
Pin, rename, search and delete chats, or export them as Markdown or JSON from the ⋯ menu.
Stop a reply mid-way and keep what was written so far. A failed request keeps your message.
Light and dark themes, saved drafts per chat, ⌘N for a new chat and ⌘K to search.
Privacy by design
AiDesk Helper talks to one thing: Ollama, running on the same computer. There is no cloud fallback and nothing to sign in to.
The app only listens on your own machine and only talks to a local model engine. Other websites can’t reach it.
Once a model is installed, you can chat with no internet connection at all. On a plane, in a cabin, anywhere.
File contents read for a reply go only to your local model and are not saved in the chat history.
Desktop app · Project folder
Open a project and AiDesk knows its files and its AGENTS.md instructions. Ask normally and it reads, searches and, when you ask, edits files for you.
Get started
AiDesk Helper runs on macOS with Apple Silicon today. Windows support is planned.
Ollama runs the AI models on your computer. It’s free.
ollama.com/downloadDownload the ZIP, double-click it and move AiDeskHelper to your Applications folder. The first time you open it, macOS says it can’t verify the app: click Done, then openSystem Settings ▸ Privacy & Security and click Open Anyway. You only do this once.
Download for macOSApple Silicon (M1 or later) · ZIP, about 120 MB.
Open Settings ▸ Models and install one. qwen3.5:9b (7.6 GB) is a good first pick; the first model you install is selected for you.
FAQ
Yes. There is no subscription, no account and no API key. You need your own computer and a model from Ollama, which is free too.
Not for chatting. The app talks only to Ollama on your own device, and Ollama goes online only to download a model when you choose to install one. The one exception is an optional update check, off by default: if you turn it on, the app asks aideskhelper.com whether a newer version exists, without sending anything about you or your chats.
In a database file inside the app’s data folder on your device, readable only by your user account. It is not encrypted, so anything running as your user can read it. Nothing is synced anywhere.
Any chat model available in Ollama. The app suggests qwen3.5:9b, qwen3.6:27b, qwen3-coder:30b and qwen2.5-coder:1.5b, and you can type any name:tag from the Ollama library. Bigger models need more memory.
Only in the desktop app, only inside the project folder you open, and only when you ask for it or a coding request needs it. Hidden paths such as .git and .env, credential files and symlinks are refused. Every reply that changed something ends with a Project changes list, and deletions go to the Trash.
macOS on Apple Silicon today. Windows support is planned.
AiDesk Helper is not notarized by Apple yet, so the first time you open it macOS says it cannot verify the developer. Click Done, open System Settings ▸ Privacy & Security and click Open Anyway next to the AiDesk Helper message. After that it opens normally.