shd-music
Enables searching and ripping high-quality audio from Deezer using an ARL cookie, with automatic decryption, tagging, and local library publication.
README
rip mcp server
A Model Context Protocol server (JSON-RPC over stdio, built for Bun) that rips high-quality audio from Deezer using an ARL cookie and publishes the finished files into the local music library that backs, mainly for navidrome, but with a few tweaks you can use this anywhere to rip music automatically.
Recommended to use this with Poke
Use lucida to rip from soundcloud, rest you can rip using deezer as it has most if not all of the released music.
Because the server runs on the same host as the music site, "upload" is just a local move/copy into the library directory no SSH/SFTP/S3 required.
Tools
| Tool | Arguments | Description |
|---|---|---|
search_music |
query: string, limit?: number |
Search the Deezer catalogue. Returns tracks + URLs. |
rip_track |
track_url: string, upload?: boolean |
Download → Blowfish-decrypt → tag a track. With upload (default true) it publishes into the library. |
upload_file |
local_path: string |
Publish an existing local file into the library. |
How it works
- Search / metadata use Deezer's public REST API (no auth).
- rip_track authenticates the ARL against the private
gw-lightgateway, requests a stream token, and asksmedia.deezer.comfor an encrypted CDN URL (FLAC with automatic fallback to MP3 320/128). - The encrypted payload is decrypted with the
BF_CBC_STRIPEscheme every third 2048-byte block is Blowfish-CBC encrypted. We decrypt natively via Bun'snode:cryptobf-cbccipher (src/mcp/deezer/crypto.ts). - Tags + cover art are embedded best-effort with ffmpeg (preferred) or metaflac. If neither binary is installed the file is still produced, just untagged.
- The result is moved/copied to
MUSIC_LIBRARY_DIRasAlbum Artist/Album/NN - Title.flac.
The download backend is pluggable (src/mcp/backends): deezer (native,
default) or lucida for a self-hosted lucida-flow instance.
Configuration
All via environment variables (see .env.example):
| Var | Default | Notes |
|---|---|---|
MUSIC_BACKEND |
deezer |
deezer or lucida |
DEEZER_ARL |
— | Required for the deezer backend |
MUSIC_QUALITY |
FLAC |
FLAC | MP3_320 | MP3_128 |
MUSIC_LIBRARY_DIR |
./library |
Publish destination |
DOWNLOAD_DIR |
OS temp dir | Working dir for decrypt/tag |
UPLOAD_MODE |
move |
move or copy |
EMBED_TAGS |
true |
Toggle ffmpeg/metaflac tagging |
LUCIDA_API_URL / LUCIDA_API_KEY |
— | Only for the lucida backend |
MCP_LOG_LEVEL |
info |
stderr verbosity |
Running
bun run mcp # production
bun run mcp:dev # watch mode
Register it with an MCP client:
{
"mcpServers": {
"rip-music": {
"command": "bun",
"args": ["run", "src/mcp/index.ts"],
"cwd": "/path/to/where/you/cloned"
}
}
}
The server logs to stderr; stdout is reserved for the JSON-RPC protocol. For tagging + embedded artwork, install
ffmpeg(recommended) orflac(metaflac) on the host.
Note
Use this only with your own Deezer account/ARL and for content you're entitled
to download. The ARL is a credential — keep it in .env (git-ignored), never
commit it.
Preview/Demo
https://github.com/user-attachments/assets/df363e87-2d62-4d06-ac3b-353ceccd2e6a
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。