Compare
Herdbell vs a DIY notification script for Claude Code
You can get Claude Code to run a command when it finishes or asks for permission, and have that command show a notification. Plenty of people do. Herdbell is built on the same idea, so the question is what you gain from an app instead of a script.
Side by side
| DIY hook script | Herdbell | |
|---|---|---|
| Alerts you when it finishes or asks | Yes, once you write it | Yes, by default |
| Answer from the alert | No: you go back to the terminal | Yes: Return to allow, D to deny, or type a reply |
| More than one agent | One script per agent, each set up by you | Claude Code, Codex, Cursor and Antigravity |
| A list of every chat and its status | No | Yes, with search and history |
| Setup | You write, test and debug it | Connect in one click; settings are backed up first |
| Upkeep | Yours when an agent changes | Updates ship with the app |
| Cost | Free | One payment |
Both build on hooks, the small commands an agent runs at set moments. If you want to understand them, read Claude Code hooks, explained.
When the script is the right call
- You only want a ping, and you are happy to switch to the terminal to answer.
- You use one agent in one terminal.
- You enjoy tuning it, or you are not on a Mac.
When Herdbell saves you the work
- You run several chats or several agents, and want one place that shows which is waiting.
- You want to answer a permission request without leaving what you are doing.
- You would rather not maintain a script every time an agent changes.
Common questions
Is a hook script safe?
It can be, if it is small and you know what it runs. A script that approves things automatically is risky. Herdbell never approves for you: it shows the request and sends back your answer.
Can I use a script and Herdbell together?
Yes. They both listen to the same agent hooks, so a personal script can run beside Herdbell.