Claude Code

rippy plugs into Claude Code as a PreToolUse hook on the Bash tool. Every shell command Claude Code wants to run is piped to rippy first; rippy approves, asks, or blocks.

One-line setup

The fastest path is to let rippy edit settings.json for you:

rippy setup claude-code

That’s it — rippy writes the hook stanza into ~/.claude/settings.json. Run it from inside a repo to install the hook at the project level instead of globally.

Manual setup

If you prefer to edit settings.json by hand, add this block to ~/.claude/settings.json (or .claude/settings.json inside a repo for project-scoped setup):

{
  "hooks": {
    "PreToolUse": [
      {
        "matcher": "Bash",
        "hooks": [
          { "type": "command", "command": "rippy --mode claude" }
        ]
      }
    ]
  }
}

What happens on each call

1. Claude Code serializes the Bash tool-use payload to JSON and pipes it to rippy.
2. rippy first checks Claude Code’s own permissions.allow / permissions.deny / permissions.ask rules in ~/.claude/settings.json as a pre-analysis step.
3. If nothing matches there, rippy parses the command into an AST, consults your .rippy.toml config, and applies its safety model.
4. rippy prints a JSON verdict on stdout. Exit code 0 means allow, 2 means ask / deny, 1 means an internal error (rippy fails open on internal errors — see the FAQ).

Claude Code permissions are honored automatically

You do not need to duplicate Claude Code’s permissions.allow / permissions.deny / permissions.ask rules into your .rippy.toml. rippy reads them directly from ~/.claude/settings.json as a separate pre-analysis step. Edit them in one place, keep both rippy and Claude Code in sync automatically.