A Mac Agent Living in the Notch
zack40xx · reddit · 2026-07-16
The author built a Mac app where an agent "living behind the notch" automatically routes natural language to either chat or action.
Core design features:
- Routing over mode switching: The system automatically determines whether to "chat" or "act," with manual Do / Ask overrides if it guesses wrong.
- Pre-write review: Before modifying reminders, to-dos, or calendars, a review card pops up and executes only after user confirmation.
- Exemptions for open-only actions: Actions that don't alter state execute immediately.
- Voice wake word gating: Local recognition, a red dot indicator, and wake words prevent accidental triggers.
- Intercepting other agents' confirmations: For example, when Claude Code or Codex requests permissions in the terminal, the notch area can also display Allow / Deny options and jump back to the respective terminal.
The author clarifies that it runs on the user's own Claude subscription via Claude Code, requiring no API key and no intermediate server. The final open question posed is: for "write actions," should review cards always be the default, or should repeated actions gradually earn auto-approval.
More from coding & agent
- A roundup of AI agents and MCP resources, including how to evaluate agents — _jaydeepkarale · 2026-07-21
- A full course shows how to build and deploy an AI agent with OpenAI and LangChain — _jaydeepkarale · 2026-07-21
- A beginner guide to AI agents points readers to a Stanford webinar — _jaydeepkarale · 2026-07-21
- A practical guide on how to evaluate AI agents — _jaydeepkarale · 2026-07-21
- MCP is headed toward easier scale, event-driven extensions, and workable file uploads — EricBuess · 2026-07-21
- Developers debate the missing composition model for AI agents — threepointone · 2026-07-21