Communitygithub.com

kk43994/kkclaw

Give Your AI Agent a Desktop Body — Setup Wizard, 14 Emotion Spheres, Voice Cloning, Lyrics Window, Doctor Self-Check, Cross-Platform (Windows + macOS)

What is kkclaw?

kkclaw is a desktop AI companion skill that gives your AI agent a tangible, interactive form, designed for platforms like Claude Code, Cursor, or Codex. It renders a customizable, emotionally expressive sphere using pure HTML/CSS/JS with CSS radial-gradient effects for a glass-like aesthetic. Key features include a 3-minute Setup Wizard, local API key encryption via safeStorage, voice cloning with MiniMax, CosyVoice, and Edge TTS with failover, 14 distinct emotional states with eye micro-expressions and dynamic glow effects, a Doctor self-diagnosis module for Gateway/model/TTS/port/logs, and real-time WebSocket synchronization with OpenClaw Gateway. It supports Windows and macOS with desktop integration features like window pinning, mouse-through, drag-follow, and a lyrics-style subtitle window, making it a lightweight, reliable 7x24 AI companion.

Works with~Claude Code~Codex CLI~Cursor
npx skills add kk43994/kkclaw

Ask in your favorite AI

Open a new chat with this agent skill pre-loaded.

Documentation

kkclaw — Desktop Lobster AI Companion

Give your AI agent a desktop body: a visual sphere + voice interaction + emotional expression.

Features

  • 🧙 Setup Wizard — First-run configuration guide, 3 minutes to get started
  • 🔐 safeStorage Encryption — API keys stored locally encrypted, no more plaintext
  • 🎭 Persona Customization — Automatically generates AGENTS.md / SOUL.md / USER.md / HEARTBEAT.md
  • 🎵 Audio System — MiniMax Voice Cloning + CosyVoice + Edge TTS triple fallback, never silent
  • 👀 Emotional Expression — 14 emotions + multiple eye micro-expressions + dynamic glow effects
  • 🩺 Doctor Self-Check — One-click diagnosis of Gateway / model / TTS / ports / logs / lyrics window
  • 🎛️ Model Management — Provider CRUD, model add/delete, latency testing, quick presets
  • 🧠 Gateway Sync — Real-time WebSocket synchronization with OpenClaw Agent state, millisecond response
  • 💪 7x24 Reliability — Auto-recovery, health score monitoring, Switch Logger
  • 🖥️ Lightweight Rendering — Pure HTML/CSS/JS sphere, CSS radial-gradient glass texture, no framework dependency
  • 🎤 Voice Broadcast — Automatic conversion of agent replies to speech, supports emotion detection and pause markers
  • 🖱️ Desktop Integration — Window pin on top, mouse-through, drag-follow, lyrics-style subtitle window

Tech Stack

  • Electron (desktop container)
  • OpenClaw Gateway (WebSocket communication)
  • MiniMax Speech / CosyVoice / Edge TTS (speech synthesis)
  • CSS Animation + requestAnimationFrame (sphere animation)

Installation

git clone https://github.com/kk43994/kkclaw.git
cd kkclaw
npm install
npm start

Or download the installer:

Configuration

Edit pet-config.json to customize:

  • TTS engine and voice
  • Sphere color and size
  • OpenClaw Gateway connection
  • Emotion mapping rules
  • Setup Wizard initial guide flow

Directory Structure

kkclaw/
├── main.js              # Electron main process
├── index.html           # Sphere rendering UI
├── lyrics.html          # Lyrics subtitle window
├── setup-wizard.html    # Setup Wizard interface
├── setup-wizard.js      # Setup Wizard logic
├── gateway-client.js    # Gateway communication client
├── smart-voice.js       # Voice scheduling system
├── templates/           # Agent template files
├── voice/               # TTS engine modules
├── utils/               # Utility modules
├── scripts/             # Build scripts
└── docs-dev/            # Development documentation

Links

Related Skills