youtube-mcp
MCP server for YouTube Data API v3 enabling video search, channel lookup, trending content, comments, and playlist management.
README
YouTube MCP Server
Model Context Protocol (MCP) server for YouTube Data API v3. Provides tools for searching videos, channels, playlists, getting trending content, video details, comments, and more.
Features
- Video Search - Search YouTube videos with advanced filters (duration, definition, region, date range, etc.)
- Channel Search - Find channels by query
- Video Details - Get comprehensive video information (statistics, content details, topics, etc.)
- Channel Details - Get channel statistics, branding, topics, and more
- Trending Videos - Get most popular videos by region and category
- Video Categories - Browse available video categories
- Channel Videos - Get recent uploads from a specific channel
- Playlist Videos - Get videos from any playlist
- Playlist Search - Search for playlists
- Video Comments - Get comments on videos
- Channel by Handle - Look up channels by @username
Installation
git clone https://github.com/PPiai/youtube-mcp.git
cd youtube-mcp
pip install -e .
Or with uv:
uv pip install -e .
Configuration
Requires a YouTube Data API v3 key. Get one from Google Cloud Console.
Copy .env.example to .env and fill in your values:
cp .env.example .env
| Variable | Default | Description |
|---|---|---|
YOUTUBE_API_KEY |
— | Required. YouTube Data API v3 key |
MCP_TRANSPORT |
streamable-http |
streamable-http (remote service) or stdio (local) |
HOST |
0.0.0.0 |
Bind host (HTTP mode) |
PORT |
8000 |
Bind port (HTTP mode) |
Transports
This server supports two transports, selected by MCP_TRANSPORT:
Streamable HTTP (remote service — default)
Runs as a persistent HTTP service on a single port (default 8000).
| Endpoint | Method | Purpose |
|---|---|---|
/mcp/ |
POST |
MCP streamable-HTTP endpoint (JSON-RPC) |
/health |
GET |
Health check — returns 200 {"status":"ok"} |
MCP_TRANSPORT=streamable-http PORT=8000 youtube-mcp
# MCP endpoint: http://localhost:8000/mcp/
# Health: http://localhost:8000/health
Note — trailing slash. The MCP endpoint is mounted at
/mcp, soPOST /mcpreturns a 307 redirect to/mcp/. Always point clients at/mcp/(with the trailing slash) to avoid clients that don't follow redirects onPOST.
Connecting an MCP client to the remote endpoint (e.g. mcp.json /
Claude Desktop style config):
{
"mcpServers": {
"youtube": {
"type": "streamable-http",
"url": "https://<your-domain>/mcp/"
}
}
}
Quick test with curl:
# Health
curl https://<your-domain>/health
# Initialize (note: streamable-HTTP replies with an SSE stream)
curl -X POST https://<your-domain>/mcp/ \
-H "Content-Type: application/json" \
-H "Accept: application/json, text/event-stream" \
-d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-06-18","capabilities":{},"clientInfo":{"name":"curl","version":"1.0"}}}'
# Call a tool
curl -X POST https://<your-domain>/mcp/ \
-H "Content-Type: application/json" \
-H "Accept: application/json, text/event-stream" \
-d '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"search_videos","arguments":{"query":"ai news","max_results":2}}}'
⚠️ Security. The HTTP endpoint has no built-in authentication — anyone who can reach the URL can issue requests and consume your YouTube API quota. Put it behind an authenticating reverse proxy / network policy, or keep it on a private network. Do not expose it publicly without protection.
stdio (local)
Launched on-demand by a local client (e.g. via uvx):
mcp_servers:
youtube:
command: "uvx"
args: ["youtube-mcp"]
env:
YOUTUBE_API_KEY: "your_api_key_here"
MCP_TRANSPORT: "stdio"
timeout: 120
connect_timeout: 60
Deploy (EasyPanel / Docker)
Build and run as a container:
docker build -t youtube-mcp .
docker run -p 8000:8000 -e YOUTUBE_API_KEY="your_api_key_here" youtube-mcp
On EasyPanel, set YOUTUBE_API_KEY as a service environment variable (not a
build-arg), expose port 8000, and point the MCP client to
https://<your-domain>/mcp.
Available Tools
| Tool | Description |
|---|---|
search_videos |
Search videos with filters |
search_channels |
Search channels |
get_video_details |
Get detailed video info |
get_channel_details |
Get detailed channel info |
get_trending_videos |
Get trending videos by region |
get_video_categories |
Get video categories |
get_channel_videos |
Get channel's recent videos |
get_playlist_videos |
Get playlist videos |
search_playlists |
Search playlists |
get_video_comments |
Get video comments |
get_channel_by_handle |
Get channel by @handle |
Example Queries
- "Search for videos about AI news from the last week"
- "Get trending videos in Brazil (BR)"
- "Find channels about programming tutorials"
- "Get details for video dQw4w9WgXcQ"
- "Get comments on a viral video"
- "Find playlists about machine learning"
Development
# Install dev dependencies
pip install -e ".[dev]"
# Run the server locally over HTTP and smoke-test it
MCP_TRANSPORT=streamable-http youtube-mcp &
curl http://localhost:8000/health
No automated test suite yet —
pytestis wired up in thedevextra but there are no tests under atests/directory. Contributions welcome.
License
MIT
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。