Archivarix Tube Search

**Archivarix Tube Search** is a search engine for archived YouTube videos, including content that has been deleted, made private, or region-blocked on the original platform. The index combines Wayback Machine snapshots, Common Crawl, the YouTube Meta 2019 dump, and proprietary crawlers — covering **1.5+ billion videos** archived since 2005.

### Key features

- **Full-text search** across video titles, channel names, descriptions, and tags
- **Deleted-video recovery** using multi-source archive reconstruction (Wayback Machine + Common Crawl + Meta 2019)
- **Thumbnail and metadata retrieval** from web archives — see what was originally there, even when the video is gone
- **AI-generated summaries** for archived videos (Gemma 4 27B), tailored to video length
- **Automatic transcripts** via self-hosted ASR pipeline (Qwen3-ASR), with downloadable .srt/.vtt subtitles
- **MCP server** for AI agents — connect Claude Desktop, Cursor, or any MCP-compatible client and let your model search the archive directly
- **API access** for developers, with usage-tier pricing
- **Channel discovery** — find every video ever published by a channel, including deleted uploads

### How it works

1. Enter a video title, channel name, keyword, or paste a YouTube URL
2. The engine searches across multiple archive sources in parallel
3. Recovered metadata appears with thumbnails, transcripts, and AI summaries where available

### Use cases

- **Journalists and researchers** — recover deleted evidence and verify removed claims
- **Content creators** — find lost source material for reuploads or commentary
- **Digital preservationists** — track what disappears from YouTube over time
- **Lost media communities** — search for forgotten or removed videos by name or era
- **AI developers** — feed archived YouTube content into agents via the MCP server

### Pricing

- **Free tier**: full-text search and metadata access
- **Plus**: higher rate limits, AI summaries, transcript downloads
- **Plus API**: programmatic access and MCP server access

Open source: the MCP server wrapper is published as [`tube-search-mcp`](https://github.com/Archivarix-com/tube-search-mcp) (MIT). The core backend is proprietary.

[Try it free →](https://tube.archivarix.net)