Tracker

Tracker is a native macOS utility for local activity context and permission-aware workflows.

It focuses on the part most AI productivity tools try to skip: what can the app actually see, what did the user approve, and how do you make that state obvious without turning onboarding into a support ticket generator?

What it explores

• Native Swift flows for screen, activity, and system permission states.
• Clear onboarding that reflects the real macOS state instead of guessing.
• Local-first workflow signals that can support future productivity tools.
• Permission-aware UX for assistants that need context without being creepy.
• A calmer foundation for desktop products that depend on system access.

Why this project exists

Useful agents need context. Good products need trust. Tracker sits in that gap and treats permissions as product design, not boilerplate.

On macOS, permissions are not a detail you can hide. Screen recording, accessibility, input monitoring, notifications, and local context all have user-facing implications. If the product guesses wrong, the user loses trust. If the onboarding is unclear, the product feels broken before it has a chance to help.

Product story

Tracker is not trying to be a giant analytics product. It is a foundation for understanding local activity state in a way that could support future desktop agents. The important pieces are native behavior, visible permission state, and clear feedback when the app needs something from the user.

That makes it related to Elson, but not identical. Elson is the voice assistant surface. Tracker is more about the local context substrate: what is visible, what is allowed, and how a desktop tool can reason about state safely.

Recent direction

Recent work tightened onboarding and made the UI line up with what macOS actually allows. That sounds small, but it is exactly the kind of product detail that determines whether a local utility feels trustworthy.

The next useful work would be clearer state visualization, sharper permission recovery flows, and a stronger bridge between local activity signals and assistant workflows.

Why it matters

AI desktop tools often jump straight to model behavior. Tracker starts earlier: before an assistant can help with context, it needs a safe way to know what context exists. That is the unglamorous layer that makes more ambitious local agents possible.

Public boundary

This page describes product direction and permission-aware UX only. It does not expose private activity data, screen contents, local logs, or user-specific context.