YouTube Transcripts

You can fetch YouTube video transcripts through an MCP server that exposes a dedicated tool for retrieving transcripts in multiple languages and formats. This server is designed to work with MCP-compatible clients, enabling LLMs to access transcripts securely and efficiently for downstream processing.

How to use

To use the YouTube transcripts MCP server, connect your MCP client to the server as a local (stdio) endpoint. Invoke the fetch_youtube_transcript tool by providing the video ID, language code, and desired output format. The tool returns either plain text or JSON transcripts that your application can parse and feed into your LLM workflows. If you already have an MCP client configured, you can start querying transcripts from your prompts or automation flows just like you would with other MCP-enabled tools.

How to install

Prerequisites: ensure you have the MCP tooling installed, and you can run the uv manager for project setup.

# Step 1: Install uv if you haven't already
# (follow your platform's instructions for uv installation)

# Step 2: Clone the YouTube transcripts MCP server repository
git clone https://github.com/PraveenKishore/mcp-server-youtube.git
cd mcp-server-youtube

# Step 3: Create and install dependencies using uv
uv sync

# Step 4: Activate the virtual environment
# Linux/macOS
source .venv/bin/activate

# Windows
.\venv\Scripts\activate

Additional setup and testing

The server is configured to run as an MCP stdio server using the uv tool. The following configuration snippet shows the MCP connection you should expose to your MCP client.

{
  "mcpServers": {
    "youtube_transcripts": {
      "command": "uv",
      "args": [
        "directory",
        "/ABSOLUTE/PATH/TO/PARENT/FOLDER/mcp-transcripts/src",
        "run",
        "server.py"
      ]
    }
  }
}

Testing the MCP server

1. Testing the MCP server functionality only: start the MCP inspector so you can view and invoke exposed tools.

mcp dev src/server.py

Notes on usage with MCP clients

Configure your MCP client with the provided stdio configuration to connect to the local server. Use the fetch_youtube_transcript tool by specifying the video ID, language code, and format. This enables you to fetch transcripts on demand within your LLM workflows.

Available tools

fetch_youtube_transcript

Fetches YouTube video transcripts for a given video ID, language code, and format, returning either plain text or JSON suitable for downstream processing by LLMs.