Claude Code Hooks Builder

A hook is a command Claude Code runs automatically at moments like “waiting for your approval” or “finished responding”. Build the config here, copy it, and paste it into ~/.claude/settings.json — that is all it takes. Pick the events you care about, write one command, and copy a settings.json snippet that Claude Code accepts as-is. Every one of the 33 hook events is here with its matcher values, plus the statusLine setting. If you need the field-by-field details, see the hooks reference.

What you're generating

A hook is a shell command Claude Code runs at a lifecycle event, with a JSON payload on stdin (session_id, hook_event_name, cwd, and per-event fields such as tool_name or notification_type). The shape the builder produces is:

{
  "hooks": {
    "<EventName>": [
      {
        "matcher": "<optional filter>",
        "hooks": [ { "type": "command", "command": "<your command>", "timeout": 600 } ]
      }
    ]
  }
}

• matcher filters which occurrences fire the hook: tool names for tool events, notification types for Notification, and so on. Plain names separated by | match exactly; anything with regex characters is treated as a JavaScript regex. Leave it empty to match everything.
• One command for all events is the common pattern — read hook_event_name from stdin and branch inside the script. The builder writes the same handler under each selected event.
• Exit code 2 blocks the action on events that support it (PreToolUse, UserPromptSubmit, PermissionRequest…), with stderr shown as the reason. Any other exit code lets it proceed.
• statusLine is a separate top-level key with a single slot. Its command gets context_window.used_percentage, rate_limits and model — data that hooks never receive.