Transcribe YouTube videos for LLM chat apps via MCP.
Example prompt: Summarize https://www.youtube.com/watch?v=uB9yZenVLzg
Install
Requires uv. Add this to your MCP client's config — works for any MCP-compatible app (Claude Desktop, Claude Code, Cursor, Windsurf, VS Code, Zed, …):
{
"mcpServers": {
"youtube-transcript": {
"command": "uvx",
"args": ["youtube-transcript-mcp-server"]
}
}
}Equivalent one-liner if your client wants a single command:
uvx youtube-transcript-mcp-server
Tool
transcribe(youtube_video_url: str) -> str— fetches the transcript (en/de/es/fr/ru) and prepends an ad-removal instruction for the LLM.
Develop locally
git clone https://github.com/SeanPedersen/youtube-transcript-mcp cd youtube-transcript-mcp uv venv && uv pip install -r pyproject.toml && source .venv/bin/activate python mcp_server.py
Point your MCP client at the local checkout instead of PyPI:
{
"command": "uv",
"args": [
"run", "--with", "fastmcp", "--with", "youtube-transcript-api",
"fastmcp", "run", "/absolute/path/to/youtube-transcript-mcp/mcp_server.py"
]
}