rullerzhou-afk/clawd-on-desk

A pixel desktop pet that watches Claude Code, Codex, Cursor & other AI coding agents — so you don't have to.

What is Clawd on Desk?

Clawd on Desk is a tiny desktop‑pet application (a pixel‑art crab, cat or cloud‑shaped creature) that watches the activity of your AI‑coding assistants (Claude Code, Copilot, Gemini, Cursor, etc.) and reacts in real time. When the assistant is thinking, typing, running tools, spawning sub‑agents, or asking for a permission, Clawd animates accordingly – sleeping when you’re idle, grooving when a sub‑agent runs, popping a permission‑bubble you can click to approve/deny, and celebrating when a job finishes.

It works on Windows 11, macOS and Linux and can be installed from pre‑built installers (x64/ARM64) or built from source with Node.js.

Core capabilities

Category What it does
Multi‑agent support Hooks for dozens of AI‑coding agents (Claude Code, Codex CLI, Copilot CLI, Gemini, Cursor, Qwen, CodeWhale, Pi, OpenClaw, etc.). You can also register a custom HTTP‑based agent.
Live animation 12 animated states (idle, thinking, typing, building, sub‑agent groove, juggling, error, happy, notification, sweeping, carrying, sleeping). Eye‑tracking follows the cursor; double‑click pokes the pet; drag‑and‑drop lets you reposition it.
Permission bubbles When an agent requests a permission, Clawd can show a floating card with Allow / Deny / Always buttons, plus hotkeys (Ctrl+Shift+Y/N). Modes let you auto‑approve, ask every time, or only for tool‑type requests.
Remote approvals Forward permission requests to Telegram or Feishu/Lark; Slack can receive read‑only notifications (task done, errors, permission prompts).
Session dashboard & HUD Right‑click → Open Dashboard to see live sessions, quotas, and jump to the corresponding terminal window. A compact HUD can stay near the pet.
Mobile companion (PWA) A token‑gated LAN‑only web app mirrors the desktop state on your phone (read‑only).
System niceties Click‑through transparent areas, remembers position, single‑instance lock, auto‑start via Claude Code hook, Do‑Not‑Disturb mode, sound effects, system‑tray menu, auto‑update via GitHub releases, and UI localisation (English, Chinese, Korean, Japanese, Portuguese, Spanish).

How to get started

  1. Download the installer for your OS from the Releases page (or clone the repo and run npm install && npm run build).
  2. Run the app – a pixel pet appears on your desktop.
  3. Open Settings → Agents and click the install button for the AI‑coding tool you use (e.g., npm run install:gemini-hooks). The command creates the appropriate hook file (~/.gemini/settings.json, ~/.copilot/hooks/hooks.json, etc.) so the agent reports its lifecycle events to Clawd.
  4. Watch the pet animate as you code. When a permission request shows up, click the bubble or use the hotkeys to respond.
  5. (Optional) Enable Remote Notifications in Settings → Remote to push permission requests to Telegram/Feishu or get Slack alerts.
  6. (Optional) Turn on Mobile / PWA to view the live state on your phone.

Who might like this?

  • Developers who spend long periods using AI‑driven coding assistants and want a visual cue that the background work is still progressing.
  • Teams that need quick, glanceable permission handling without switching to a terminal.
  • Anyone who enjoys a playful desktop companion that reflects the AI‑agent’s activity.

Where to learn more

  • Setup guide – docs/guides/setup-guide.md
  • Custom agent integration – docs/guides/custom-agent-http.md
  • Slack notification docs – docs/guides/slack-notifications.md
  • Mobile companion repo – https://github.com/Bynlk/clawd-on-mobile

Quick reference

Feature Detail
Supported agents Claude Code, Codex CLI, Copilot CLI, Gemini CLI, Antigravity, Cursor, CodeBuddy, WorkBuddy, Kiro, Kimi‑CLI, Qwen, ZCode, CodeWhale, Reasonix, opencode, MiMo, Pi, OpenClaw, Hermes, Qoder, QoderWork, QwenWork, DeepSeek Harness
Platforms Windows 11 (x64/ARM64), macOS, Ubuntu/Linux
Languages UI: English + 6 translations; themes can be custom‑made (pixel‑art atlases)
License (not specified in README – check repository)
Latest version Shown by GitHub release badge

Clawd on Desk turns the invisible work of AI coding agents into a friendly, interactive desktop experience, giving you visual feedback, easy permission handling, and optional remote notifications—all while looking cute.

Related

  • Project
    HermannBjorgvin/Clawdmeter

    Clawdmeter is an ESP‑32‑based desk gadget that shows your Claude Code usage (session and weekly percentages) on a small OLED screen. A Python daemon on macOS/Linux/Windows polls the Claude API, extracts usage from response headers, and sends the data over BLE to the device. The firmware (LVGL UI) displays the numbers and animates a mascot whose activity reflects usage level. Two side buttons act as BLE‑HID keys (Space for voice‑mode, Shift + Tab for mode toggle). The repo provides flashing scripts, platform‑specific installers, a desktop simulator, and documentation for porting to other ESP‑32 boards.

  • Project
    m1ckc3s/claude-status-bar

    A macOS menu bar app that provides live status indicators for Claude Code, showing whether the AI is working, thinking, or awaiting user permission.

  • Project
    richhickson/claudecodeusage

    Claude Usage is a native macOS menubar app that shows real‑time Claude Code quota usage (session, weekly, monthly) and alerts you when a Claude Code session is waiting for input, letting you jump directly to the terminal tab. It reads your OAuth token from the Keychain, polls Anthropic’s usage endpoint, and uses Claude Code’s hook system for session monitoring.

  • Project
    OpenLAIR/dr-claw

    Dr. Claw is an open‑source, model‑agnostic AI research assistant that provides a unified UI/CLI for the whole research workflow—literature survey, idea generation, experiment code, analysis, paper drafting, and slide creation—driven by Claude Code, Gemini, Codex or any OpenRouter model.

  • Project
    OpenCoworkAI/open-cowork

    Open Cowork is a free, open‑source desktop app that wraps Claude‑style (and other) LLMs into a local AI‑agent workspace. With one‑click installers for Windows/macOS, VM‑level sandboxing (WSL2/Lima), a built‑in “Skills” system for generating PPTX/DOCX/PDF/XLSX files, MCP connectors for browsers/Notion, and remote control via Feishu or Slack, it lets non‑technical users give natural‑language commands that read/write files, automate GUI apps, and produce professional documents—all running locally with optional VM isolation.