AI recommendations, finished.
AI is already good at helping people discover music. The annoying part is everything after the recommendation: searching YouTube, picking the right upload, avoiding karaoke and bad edits, copying tracks one by one, and building the playlist manually.
Deepcut closes that gap. It turns a natural-language music request into a real playlist on the YouTube account you choose.
Search, matching, filtering, playlist construction, and safe playlist management happen behind the scenes.
Deepcut is built on the Model Context Protocol so the same service can work across compatible AI clients. The backend is provider-neutral; ChatGPT, Claude, Cursor, and other MCP clients can use the same hosted endpoint.
What makes it different
Deepcut does more than forward YouTube API calls. It ranks candidates, prefers official releases, rejects common bad versions, checks confidence, manages quota, deduplicates writes, and requires explicit confirmation for destructive actions.
Current status
The local product is working against a real YouTube account. Public hosted multi-user mode is the next release step, followed by Google OAuth verification, quota expansion, and submissions to supported AI directories.