youtube-live-mcp
Enables AI assistants to control YouTube Live Streaming, including creating broadcasts, managing chat, and monitoring stream health via MCP tools.
README
YouTube Live MCP
MCP server for controlling YouTube Live Streaming from Claude Code, Claude Desktop, or any MCP client.
Create broadcasts, go live, manage chat, and monitor stream health — all through the Model Context Protocol.
What it does
This server wraps the YouTube Live Streaming API v3 and exposes it as MCP tools. Combined with obs-mcp, you get fully automated live streaming from your AI assistant.
Tools
Broadcasts (8 tools)
yt-create-broadcast— Create a new live broadcastyt-list-broadcasts— List broadcasts (filter by status)yt-get-broadcast— Get broadcast detailsyt-update-broadcast— Update title, description, privacyyt-delete-broadcast— Delete a broadcastyt-bind-stream— Bind a video stream to a broadcastyt-transition-broadcast— Go live, end broadcast (testing/live/complete)yt-insert-cuepoint— Insert ad break during live
Streams (4 tools)
yt-create-stream— Create stream, get RTMP URL + keyyt-list-streams— List your streamsyt-get-stream— Get stream details + healthyt-delete-stream— Delete a stream
Chat (6 tools)
yt-list-chat-messages— Read live chatyt-send-chat-message— Post to live chatyt-delete-chat-message— Remove a messageyt-list-moderators— List chat moderatorsyt-add-moderator— Add a moderatoryt-remove-moderator— Remove a moderator
Status (2 tools)
yt-get-broadcast-status— Broadcast lifecycle stateyt-get-stream-health— Real-time stream health
Videos
yt-search-videos— Search YouTube for videos by keyword. Params:query(required),maxResults(1–50, default 10),channelId(optional, restrict to one channel),order(relevance/date/viewCount/rating/title, default relevance). Returns{ videoId, title, channelTitle, channelId, publishedAt, url, description }per result. Note:search.listcosts 100 quota units per call (vs ~1 for list/get).
Prerequisites
- Node.js 18+
- Google Cloud project with the YouTube Data API v3 enabled
- OAuth2 credentials (Desktop application type)
Setting up Google Cloud
- Go to Google Cloud Console
- Create a project (or use an existing one)
- Enable the YouTube Data API v3
- Go to Credentials → Create Credentials → OAuth 2.0 Client ID
- Application type: Desktop app
- Note your Client ID and Client Secret
Installation
Claude Desktop / Claude Code
Add to your claude_desktop_config.json:
{
"mcpServers": {
"youtube-live": {
"command": "npx",
"args": ["-y", "youtube-live-mcp@latest"],
"env": {
"YOUTUBE_CLIENT_ID": "your-client-id.apps.googleusercontent.com",
"YOUTUBE_CLIENT_SECRET": "your-client-secret"
}
}
}
}
Manual
npm install -g youtube-live-mcp
YOUTUBE_CLIENT_ID="..." YOUTUBE_CLIENT_SECRET="..." youtube-live-mcp
Authentication
On first run, the server opens your browser for Google OAuth consent. After authorizing, tokens are stored in ~/.youtube-live-mcp/tokens.json and refreshed automatically.
Required scope: https://www.googleapis.com/auth/youtube
Usage: Go Live in 7 Steps
With both youtube-live-mcp and obs-mcp installed:
1. yt-create-broadcast → Create the YouTube event
2. yt-create-stream → Get RTMP URL + stream key
3. obs-set-stream-settings → Configure OBS with RTMP details
4. yt-bind-stream → Connect stream to broadcast
5. obs-start-stream → Start sending video from OBS
6. yt-transition-broadcast → testing → validate feed
7. yt-transition-broadcast → live → YOU'RE LIVE
To end: yt-transition-broadcast complete + obs-stop-stream
Environment Variables
| Variable | Required | Description |
|---|---|---|
YOUTUBE_CLIENT_ID |
Yes | OAuth2 client ID |
YOUTUBE_CLIENT_SECRET |
Yes | OAuth2 client secret |
API Quota
YouTube API has a daily quota of 10,000 units. Read operations cost ~1 unit, write operations cost ~50 units. Normal streaming use stays well within limits.
One exception: search.list (used by yt-search-videos) costs 100 units per call — about 100 searches per day on the default quota. When the quota is spent, search tools return a clear quotaExceeded message; prefer yt-list-videos (by channel) for cheap listing.
License
MIT
Built by
Runway Services — a humanistic sales and marketing agency.
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。