AgentNotch vs terminal-notifier
terminal-notifier fires a macOS banner when a command ends. AgentNotch shows always-on agent status. Here is how they differ.
❯ claude
● Wiring the license claim endpoint
⎿ Edited lib/license/store.ts (+38 −6)
⎿ Edited app/api/license/claim/route.ts (+21 −3)
⎿ Ran npm test (42 passed)
● Run npm run db:migrate?
❯ 1. Yes 2. No, tell Claude what to do
waiting for your approval
Want a beep when a long command ends? terminal-notifier does exactly that, for free. It's a command-line tool that posts a native macOS banner from a script. Wrap your agent command so it calls terminal-notifier on exit and a banner lands when the run finishes. For one agent you're waiting on, that's often the whole job.
The trouble is a banner is a single scripted event. It fires once, says whatever you told it, and vanishes. It doesn't know an agent stalled, stopped for approval, or that a second session is still grinding. Run a few and you get a stack of near-identical banners, each gone in a second, none saying which session to open.
What a banner can't hold
AgentNotch watches your Claude Code, Cursor, and Codex sessions and keeps their status in the menu bar the whole time. It flags the one that needs you, shows a context gauge and plan usage per session, and clicks straight through to it. Keep your banners if you like them; AgentNotch covers the layer they were never built for.
Side by side
| terminal-notifier | AgentNotch | |
|---|---|---|
| What it does | Posts a single macOS notification from a script when a command you wired up finishes. | Watches your agents and keeps their live status in the menu bar as they run. |
| Where it lives | In your shell config. You script each notification yourself. | In the notch and menu bar, visible while you work in any other app. |
| Tells you an agent needs you? | Only if you scripted that exact event. It fires once and forgets. | Yes. It flags the one session waiting on approval, stalled, or done. |
| Shows context and plan usage? | No. A notification is a one-line message, not a live gauge. | Yes. A context gauge and plan usage, per session, at a glance. |
| Knows one agent from another? | No. Every banner looks the same unless you write the text yourself. | Yes. It reads Claude Code, Cursor, and Codex and labels each session. |
| Price | Free and open source. | One-time $9.99. 14-day free trial, no card. |
Questions
- Can terminal-notifier tell me when Claude Code finishes?
- Yes, if you wrap the command so it calls terminal-notifier on exit. It posts a macOS banner when that command ends, which works well for a single run you are waiting on.
- Why isn't a notification enough for multiple agents?
- A notification is a one-shot event, not a status. Run three agents and you get identical banners with no sense of which session or how much context is left, while AgentNotch keeps that state live in the menu bar.
- Does AgentNotch replace terminal-notifier?
- No, terminal-notifier sends the pings you script and AgentNotch shows the ongoing status you can't. They sit at different layers, so keep both if scripted banners are already useful to you.
- Do I have to configure anything per command?
- No, AgentNotch reads your Claude Code, Cursor, and Codex sessions on its own, with no per-command wiring. It stays 100% local and never reads your code or API keys.
Keep reading
See which agent needs you.
Try AgentNotchfree for 14 days, no card. Watch Claude Code, Cursor, and Codex from your Mac's menu bar.
Updated 2026-07-17