YouTube Downloader MCP Server 📺

This project is an MCP (Multi-Channel Protocol) server that allows Claude (or any MCP-compatible AI agent) to download YouTube videos directly to your computer's Downloads folder. It is designed for seamless integration with Claude's tool-calling capabilities.

Features

• Exposes two MCP tools:

  • download_youtube_video(url: str): Starts a YouTube video download in the background and returns a download ID.

  • check_download_status(download_id: str): Lets you check the status of a download (in progress, completed, or error).

• Downloads are saved to your system's Downloads folder.

• Fully async/background operation.

Related MCP server: yt-dlp-mcp-server

Installation

1. Clone the repository:

   git clone <your-repo-url>
   cd youtube-downloader-mcp

2. Install Poetry (if you don't have it):

   pip install poetry

3. Install dependencies:

   poetry install

4. Install FFmpeg

   Check and follow the instructions to install this tool using this website (https://ffmpeg.org/). The yt_dlp needs FFmpeg to merge the audio and video files into a single mp4 file.

Running the MCP Server

Run the server using Poetry:

poetry run python endpoints.py

You should see a message like:

Running YouTube Downloader MCP server

Using with Claude

1. Add this server as a tool in Claude's settings (or use the Claude desktop app's "Add MCP server" feature).

   • Update your claude_desktop_config.json adding this MCP Server configuration:

   •   {
           "mcpServers": {
               "youtube-dowloader": {
               "command": "poetry",
               "args": [
                   "--directory",
                   "<your_project_path>",
                   "run",
                   "python",
                   "endpoints.py"
               ]
               }
           }
       }

   • For more info, please check this link https://modelcontextprotocol.io/quickstart/user

2. Claude will automatically discover the following tools:

   • download_youtube_video(url: str)

   • check_download_status(download_id: str)

3. Example workflow:

Running Tests

To run the unit tests for the YoutubeDownloader service, use:

poetry run python -m unittest tests/test_services.py

All tests are located in the tests/ directory. These tests use mocks, so no real downloads or system calls are made.

Notes

• FFmpeg is required for video merging. If you don't have it, download from ffmpeg.org and ensure it's in your PATH.

• All downloads go to your system's Downloads folder.

• This server is designed for use with Claude, but any MCP-compatible client can use it.

Troubleshooting

• If you see errors about non-JSON output, make sure you are running the latest code (all yt-dlp logs are redirected to stderr).

• If you get ffmpeg not installed errors, install FFmpeg and restart the server.

• If you have issues with status not updating, ensure only one server instance is running at a time.

License

MIT