mcp-youtube 0.2.0

MCP server wrapping yt-dlp for video/audio/subtitle download via FastMCP

MCP‑YouTube 🚀

MCP‑YouTube exposes the power of yt‑dlp as a Model Context Protocol (MCP) server – download videos, playlists, audio, subtitles and more through a single JSON call.

---

✨ Highlights

• Drop‑in server – start it and invoke high‑level tools immediately
• Works with any MCP client: CLI, Cursor (VS Code), Claude Desktop, HTTP fetchers, ChatGPT plug‑ins & more
• Rich logs – progress, speed, ETA and errors stream back to the caller
• Fully async – multiple downloads at once without blocking

---

🗒️ Requirements

Component	Why it's needed	Install (macOS / Ubuntu / Windows)
Python 3.8+	Runs the MCP‑YouTube server	Built‑in / brew install python / sudo apt install python3 / https://www.python.org/downloads/
ffmpeg	Used by yt‑dlp to merge video & audio, extract audio, convert thumbnails	brew install ffmpeg / sudo apt install ffmpeg / choco install ffmpeg
yt‑dlp	Core downloader library	Installed automatically via uv sync or pip install -e .[all]

If you use the provided Docker image, both yt-dlp and ffmpeg are already included.

---

⚡️ Quick Start

# 1. Create & activate a virtualenv (using uv)
uv venv && source .venv/bin/activate

# 2. Install project dependencies
uv sync   # reads pyproject / uv.lock

# 3. Run the server (STDIO transport)
uv run python mcp_youtube.py

The banner should show:

✔ FastMCP server "mcp-youtube" is ready (stdio)

Hot‑reload for development: uv run dev python mcp_youtube.py

---

🛠️ Available Tools

Tool	Purpose
download_video	Download a single video (select quality / resolution)
download_playlist	Download a range of items from a playlist
download_audio	Extract audio only (mp3 / aac / opus …)
download_subtitles	Fetch or embed subtitles
download_thumbnail	Highest‑quality thumbnail
get_metadata	Return video metadata as JSON

📂 Output paths

Files generated by each tool are saved in type‑specific sub‑folders under the downloads/ directory.

Type	Path	Example
Video	downloads/videos/	downloads/videos/Summer Festival [abc123].mp4
Audio	downloads/audio/	downloads/audio/Summer Festival [abc123].mp3
Subtitles	downloads/subtitles/	downloads/subtitles/Summer Festival [abc123].ja.vtt
Thumbnail	downloads/thumbnails/	downloads/thumbnails/Summer Festival [abc123].jpg

Naming template:

{title} [{id}][.{lang}].{ext}

{lang} is only appended for subtitle files.

---

🌐 Using MCP‑YouTube with different clients

Client	How to connect
FastMCP CLI	fastmcp call download_video '{"url":"…"}'
Cursor (VS Code)	Place this repo in your workspace – Cursor auto‑detects .cursorrules and spawns the server on demand
Claude Desktop	Add a mcpServers entry (see below)
Any HTTP tool (curl, Postman, JS fetch, etc.)	Start server with HTTP transport – uv run python -m fastmcp.server http mcp_youtube.py --port 8000
Python code	from mcp.client import Client (example below)
ChatGPT Plugins / LLM Agents	Register the server's schema endpoint (served automatically when using HTTP)

FastMCP CLI example

fastmcp call download_video '{
  "url": "https://youtu.be/dQw4w9WgXcQ",
  "quality": "best",
  "format": "mp4",
  "resolution": "1080p"
}'

Python client example

from mcp.client import Client

client = Client.spawn("mcp-youtube")  # auto‑spawns the process
path = client.download_video(
    url="https://youtu.be/dQw4w9WgXcQ",
    quality="best",
    format="mp4",
    resolution="1080p",
)
print("Saved to", path)

Claude Desktop config (macOS)

Create/append ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "youtube": {
      "command": "uv",
      "args": [
        "--directory",
        "/ABS/PATH/TO/mcp-youtube",
        "run",
        "python",
        "mcp_youtube.py"
      ]
    }
  }
}

---

⚖️ Legal Notice & Compliance

MCP‑YouTube is strictly a convenience wrapper around the publicly available yt‑dlp project. It is your responsibility to ensure that any content you access, download, or redistribute through this tool does not violate:

1. Copyright and neighbouring‑rights legislation that applies in your jurisdiction.
2. The YouTube Terms of Service.

By using this software you acknowledge that:

• You will only download material when you have a legal right to do so (e.g. content in the public domain, your own uploads, or where the copyright holder has granted permission). • You will not use MCP‑YouTube to circumvent technical protection measures or content access restrictions. • You assume full responsibility for any legal consequences arising from your use of this project.

No Warranty / No Liability

This software is provided "as is", without warranty of any kind. In no event shall the authors or contributors be liable for any claim, damages, or other liability arising from, out of, or in connection with the software or the use or other dealings in the software.

🐳 Docker

docker build -t mcp-youtube .
docker run -it --rm -v "$PWD:/downloads" mcp-youtube

---

🛠 Development

uv run ruff check .   # lint
uv run black .        # format
uv run mypy .         # type‑check

PRs & tests welcome!

---

📜 License

MIT

---

About MCP

The Model Context Protocol standardises how LLMs interact with external tools, resources and prompts. An MCP server (like MCP‑YouTube) advertises a JSON schema for each tool; an MCP client validates arguments before forwarding the call, ensuring safe and predictable execution.

Learn more at the official repo.