A yt-dlp GUI That Sprouted an AI Agent Control Room

What it does

OmniGet is a cross-platform desktop app built with Tauri, Rust, and Svelte. One half is a download manager and media toolbox: paste a link from YouTube, Instagram, TikTok, Udemy, Hotmart, or roughly 1,800 other sites to grab video, courses, torrents, or subtitles. It also packs 156 utility tools—Whisper transcription, PDF conversion, Anki flashcards, and a music player—into the same window. The other half is a graphical control center for AI coding agents such as Claude Code, Codex CLI, Gemini CLI, and local Ollama models, giving them a sandboxed folder, a permission gate with on-screen diffs, and one-click undo.

The interesting bit

The agent layer is more than a chat wrapper. It runs shell_exec under macOS seatbelt, checkpoints the entire folder to a shadow git before any write, and can loop a job until a test command passes. There is also a durable job queue with cron and webhook triggers, an MCP server, and an isometric “World” view where you watch agents work at virtual benches. It is an unusually deep control system for local agent execution.

Key highlights

• Downloads from 1,800+ sites via yt-dlp, plus dedicated handlers for purchased Udemy, Hotmart, and Kiwify courses.
• 156 tools in 24 categories, including transcription, PDF utilities, video conversion, and CTF tools.
• GUI for Claude Code, Codex, Gemini CLI, and local LLMs via Ollama or LM Studio; runs fully offline with local models.
• Sandboxed shell on macOS, permission diffs with allow/always/deny rules, and turn-level undo via shadow-git checkpoints.
• Durable jobs with loops, cron triggers, webhooks, and a built-in MCP server for Claude Code, Cursor, and VS Code.

Caveats

• The README is heavy on feature lists and light on architecture details; it is unclear how the Udemy and Hotmart downloaders handle authentication or DRM.
• With 156 tools, two major modes, and a 3D agent world, the surface area is vast for a single codebase.

Verdict

Developers who want a graphical yt-dlp client and a local, permissioned sandbox for AI coding agents in one place will find OmniGet ambitious and convenient. Those who prefer focused, single-purpose tools may find the scope overwhelming.