The YouTube MCP Server lets you download transcripts from YouTube videos and attach them to an ongoing MCP conversation with a timestamped transcript. It accepts both YouTube video IDs and full YouTube URLs and returns a transcript that can be consumed by an MCP client to provide context to a large language model. This makes it easy to reference exact moments in a video when discussing or querying the content. The server is designed to work with any MCP-compatible client, so you can integrate it into your existing MCP workflow and prompt pipelines.

Prerequisites:

• Python 3.8+ (or a compatible Python environment)
• Internet access to clone the repository and install dependencies

Installation steps:

1. Clone the repository: git clone https://github.com/adhikasp/mcp-youtube.git cd mcp-youtube

2. Create and activate a Python virtual environment (recommended): uv venv

   On Unix/macOS

   source .venv/bin/activate

   On Windows

   .venv\Scripts\activate

3. Install dependencies via uv (as described in the project): uv sync

4. Run the server locally to verify: python -m youtube_mcp

5. Optionally pin to a specific Python version or manage with your preferred environment manager.

Notes and tips:

• The server is accessed via MCP clients using the uvx pipeline (as shown in the usage example).
• Ensure you have network access to YouTube for transcription downloads.
• If you run into issues with dependencies, refresh the virtual environment and reinstall with uv sync.
• The transcript output includes timestamps, which helps with precise context in prompts.
• This server does not require any API keys for YouTube transcription in many cases, but some regional restrictions or YouTube changes may affect availability. If needed, check the repository for any required updates or alternative download methods.
• Since this is a Python-based MCP server, you’ll configure it in your MCP client using the uvx command as shown in the usage section.