What it does

Claude Code Haha takes the infamous March 2026 leak from Anthropic’s npm registry and rebuilds it into a Tauri 2 desktop app for macOS and Windows. You get tabbed sessions, branch-aware worktrees, side-by-side code diffs, and a permission gate for dangerous commands — all the stuff that makes terminal-only AI coding feel like driving with the hood welded shut.

The interesting bit

The project doesn’t just slap a UI on leaked code. It adds Computer Use (screen capture, mouse/keyboard control), H5 remote access via one-time tokens, and IM integrations (Telegram, Feishu, WeChat, DingTalk) so you can approve agent actions from your phone. That’s a lot of surface area for a solo side project.

Key highlights

• Multi-session workspace with project switching, terminal access, and session history in one window
• Branch/Worktree isolation — start sessions against specific branches without polluting your working tree
• Right-side changes panel shows modified files, line counts, and workspace state while you chat
• Permission flows for risky commands, tool calls, and AI clarifications, all routed through the GUI
• Multi-provider support — Anthropic-compatible APIs, OpenAI, DeepSeek, Ollama, plus web search fallback
• Remote access via H5 tokens or IM bots, plus scheduled tasks and local token-usage tracking

Caveats

• The source is a leak, and the README explicitly states “for learning and research purposes only”; the legal and ethical status is murky
• macOS Gatekeeper may block the app on first launch, requiring manual override
• Personal spare-time project with commercial sponsorships — long-term maintenance is uncertain

Verdict

Worth a look if you want Claude Code’s agentic coding with a visual safety rail and remote control. Skip it if the provenance makes you nervous or if you need enterprise-grade support.