I have been running more local AI coding agents lately: Codex, Claude Code, Gemini CLI, Cursor Agent, and a few terminal-heavy workflows. The useful part is obvious, but the day-to-day ergonomics can get messy quickly.
Each agent has its own terminal, its own waiting state, its own approval prompts, and its own history. If I look away at the wrong moment, I miss the fact that one of them is blocked waiting for a decision.
So I built Agent Island, a native macOS utility that turns the notch area into a small workspace for local AI coding agents.
It is not another coding agent. It sits next to the tools I already use and keeps their session state visible.
I wanted the state to be visible without becoming another window to manage. The top of the screen is already where I look for system status, so using it as a tiny control surface for long-running coding agents felt natural.
The Mac App Store build focuses on a sandbox-safe monitoring and review flow. The direct-web build has a 7-day trial, and Pro is a one-time $19.99 license for one Mac.
Website: [https://agentisland.lzw-glory.top/](https://agentisland.lzw-glory.top/)
App Store: [https://apps.apple.com/us/app/agent-island/id6771757996?mt=12](https://apps.apple.com/us/app/agent-island/id6771757996?mt=12)
Disclosure: I am the developer of Agent Island. I am sharing it here because I think the workflow problem is becoming common for people who run multiple local AI agents.