A native Mac app with a small local Python backend

I’m Joe, the creator of JoeBro. I chose a native SwiftUI app for the Mac interface and kept the backend small: it is Python using the standard library, bundled inside the app and started when the app launches. The app talks to that backend on the Mac over localhost. The backend has no third-party Python packages to install.

That structure keeps the local workspace in one Mac app instead of asking someone to assemble a web server and a collection of services before the first conversation. JoeBro includes chats, documents, long-term memory, tools, email, calendar, deep research and other workspace features. The Mac requirements are macOS 26 or later on Apple Silicon, Xcode 26 or later to build, and Python 3. For the downloaded release, the README notes that Gatekeeper requires a one-time manual open because the indie app is not notarized.

“Local backend” describes where the app’s own service runs. It does not mean all AI processing is necessarily on the Mac. JoeBro does not bundle a model: you can configure a local model endpoint, such as Ollama, or an API key for a hosted provider. When a hosted model is selected, prompt content goes to that provider. Cloak Mode replaces detected sensitive values before requests, including local endpoints, but is pattern-based and does not erase the model-provider distinction.

The phone companion makes the distinction clearer. JoeBro Mobile is a source release that you build with Xcode using your own bundle identity. In the documented public 1.3 setup, it pairs with JoeBro running on a Mac, which must stay open and awake. Pairing is off until enabled and uses a token; Tailscale is an option for reaching the Mac away from home. My personal always-on Raspberry Pi arrangement is separate: it is not a documented supported standalone Pi mode, and there is no Pi guide.

I’m Joe, creator of JoeBro and JoeBro Mobile. The source repositories describe the current release and setup.