Give your AI
hands on YouTube.
An open-source Model Context Protocol server that lets Claude — or any MCP client — search videos, read transcripts, and scout channels. No SDK, no glue code.
$ npx -y zubeid-youtube-mcp-server
What a session looks like
Your AI chains tools on its own: search for videos, pick one, pull the transcript, answer from the actual words spoken.
Ten tools, four services
Every tool maps to the YouTube Data API v3 — plus keyless transcript retrieval and creator-discovery filters you won't find in the raw API.
Install
Pick your client. Everything below is copy-paste ready.
Add this to claude_desktop_config.json (macOS: ~/Library/Application Support/Claude/, Windows: %APPDATA%\Claude\) or your project's .mcp.json. Extra keys are optional fallbacks — the server rotates to the next key when one runs out of quota.
{
"mcpServers": {
"youtube": {
"command": "npx",
"args": ["-y", "zubeid-youtube-mcp-server"],
"env": {
"YOUTUBE_API_KEY": "your_youtube_api_key"
}
}
}
}
Add to .vscode/mcp.json in your workspace — VS Code prompts for the key and stores it securely.
{
"inputs": [
{ "type": "promptString", "id": "apiKey", "description": "YouTube API Key", "password": true }
],
"servers": {
"youtube": {
"command": "npx",
"args": ["-y", "zubeid-youtube-mcp-server"],
"env": { "YOUTUBE_API_KEY": "${input:apiKey}" }
}
}
}
One command installs and configures the server for Claude Desktop.
npx -y @smithery/cli install @ZubeidHendricks/youtube --client claude
The container serves MCP over HTTP at /mcp with a readiness probe at /ready — drop it behind n8n, LibreChat, or anything that speaks streamable HTTP.
# build with your .env present, then:
docker build -t youtube-mcp-server .
docker run --rm -p 8088:8088 youtube-mcp-server
You'll need a free YouTube Data API v3 key from the Google Cloud Console for video, channel, and playlist tools. Transcripts work with no key at all.