Home

🐤 Less permission fatigue, more momentum. Dippy knows what’s safe to run and keeps Claude on track when plans change.

How It Works

AI requests command → Dippy analyzes → allow/ask/deny → Response to AI

Decision	Meaning
allow	Safe command, auto-approved
ask	Ambiguous, prompt user
deny	Dangerous, blocked with message

Quick Start

Install

Homebrew (recommended):

brew tap ldayton/dippy
brew install dippy

Manual:

git clone https://github.com/ldayton/Dippy.git

Configure

Add to ~/.claude/settings.json (or use /hooks interactively):

{
  "hooks": {
    "PreToolUse": [
      {
        "matcher": "Bash",
        "hooks": [{ "type": "command", "command": "dippy" }]
      }
    ]
  }
}

If you installed manually, use the full path instead: /path/to/Dippy/bin/dippy-hook

Customize (optional)

Create a config file at ~/.dippy/config or .dippy in your project root:

allow git                        # Matches git, git commit, git push, etc.
deny rm -rf "Use trash instead"  # Matches rm -rf, rm -rf /, etc.

Key Features

• Zero dependencies — Pure Python, hand-written bash parser
• Multi-platform — Claude Code, Cursor, Gemini CLI, Codex CLI
• Allowlist-based — Approves known-safe commands, asks about unknowns
• Project configs — .dippy file in project root for per-project rules
• Audit logging — Optional JSON log of all decisions

Philosophy

Dippy is not adversarial — it protects against AI mistakes, not malicious actors. It uses conservative allowlists: approve what we know is safe, ask about everything else.

Documentation

Core:

• Configuration — Config file syntax and options

Extensions:

• Afterthoughts — Post-command feedback to guide Claude
• MCP Tools — Rules for Model Context Protocol tools

Proposals:

• File Editing — Rules for file editing tools

Links

• GitHub Repository
• Security Model