yt-playlist-mcp
Enables searching YouTube and managing playlists on your own account via OAuth 2.0 authentication.
README
yt-playlist-mcp
Local MCP server for Claude Desktop that searches YouTube and creates playlists on your own account.
Tools
search_youtube(query, max_results)— video search with title, channel, duration, URLcreate_playlist(title, video_ids[], description?)— creates a private playlist and adds the videos in ordersearch_playlists(query?)— lists your own playlists (optionally filtered by title substring) so later sessions can find one by IDadd_to_playlist(playlist_id, video_ids[])— appends videos to an existing playlistlist_playlist_items(playlist_id)— lists a playlist's videos in orderremove_video(playlist_id, video_id)— removes all occurrences of a video from a playlistreorder_playlist(playlist_id, video_ids[])— puts the given videos first, in order; pass the full list to sort a whole playlistdelete_playlist(playlist_id)— permanently deletes a whole playlist
Auth is OAuth 2.0 (Desktop-app client): a one-time browser consent flow stores
a refresh token at ~/.config/yt-playlist-mcp/token.json (mode 600); after
that the server refreshes access tokens silently.
1. Google Cloud setup (one time, in the browser)
- Go to https://console.cloud.google.com/ and create a new project
(e.g.
yt-playlist-mcp). - Enable the API: APIs & Services → Library → search "YouTube Data API v3" → Enable.
- OAuth consent screen: APIs & Services → OAuth consent screen
(Google may call this "Google Auth Platform → Branding/Audience").
- User type: External
- App name / support email / developer email: anything (only you will see it)
- Scopes: you can skip adding scopes here; the app requests
https://www.googleapis.com/auth/youtubeat runtime - Publish the app (Audience → "Publish app" → confirm "In production"). Leaving it in Testing status makes Google expire the refresh token every 7 days, forcing you to re-run the auth flow weekly. Published but unverified is fine for personal use — you'll click through one "Google hasn't verified this app" warning during consent (Advanced → "Go to yt-playlist-mcp (unsafe)").
- Create credentials: APIs & Services → Credentials → Create credentials → OAuth client ID → Application type: Desktop app. Copy the Client ID and Client secret.
2. Build and authorize
npm install
npm run build
YOUTUBE_CLIENT_ID=xxx.apps.googleusercontent.com \
YOUTUBE_CLIENT_SECRET=yyy \
npm run auth
npm run auth opens your browser; sign in with the Google account whose
YouTube you want to manage, click through the unverified-app warning, and
approve. Tokens land in ~/.config/yt-playlist-mcp/token.json.
Re-run npm run auth any time to re-authorize (e.g. if you revoke access at
https://myaccount.google.com/permissions).
3. Hook up Claude Desktop
Edit ~/Library/Application Support/Claude/claude_desktop_config.json
(create it if missing) — use absolute paths, since Claude Desktop does not
launch servers from this directory:
{
"mcpServers": {
"youtube": {
"command": "/absolute/path/to/node",
"args": ["/Users/will/Projects/youtube-mcp/dist/index.js"],
"env": {
"YOUTUBE_CLIENT_ID": "xxx.apps.googleusercontent.com",
"YOUTUBE_CLIENT_SECRET": "yyy"
}
}
}
}
(which node prints the node path.) Restart Claude Desktop; the two tools
appear under the youtube server.
Quota
The YouTube Data API grants 10,000 units/day by default:
search_youtube: ~101 units per call (search 100 + videos.list 1)create_playlist: 50 units + 50 per video addedadd_to_playlist: 50 units per videoremove_video: ~51 units (lookup 1 + delete 50 per occurrence)search_playlists: 1 unit per 50 playlistslist_playlist_items: 1 unit per 50 videosdelete_playlist: ~51 unitsreorder_playlist: 50 units per video actually moved (already-in-place videos cost nothing)
So roughly 90 searches/day, or fewer if you create large playlists. Quota
errors come back as tool errors mentioning quotaExceeded.
Development
npm run build # tsc → dist/
npm start # run the stdio server directly (for inspector/debugging)
npx @modelcontextprotocol/inspector node dist/index.js # interactive test UI
Secrets never live in the repo: client ID/secret come from env vars
(.env.example documents them), tokens live under ~/.config/yt-playlist-mcp/.
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。