home / mcp / mcp youtube intelligence mcp server

MCP YouTube Intelligence MCP Server

Provides a YouTube video analysis MCP server with server-side summarization, topics, entities, and reports.

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "janghyuckyun-mcp-youtube-intelligence": {
      "command": "uvx",
      "args": [
        "mcp-youtube-intelligence"
      ],
      "env": {
        "MYI_LLM_PROVIDER": "ollama",
        "MYI_OLLAMA_MODEL": "qwen2.5:7b"
      }
    }
  }
}

You can analyze YouTube videos intelligently with this MCP server and CLI. It processes original subtitles on the server and delivers concise, token-efficient results to your LLM, enabling fast video summaries, topic insights, entity extraction, and rich reports.

How to use

Connect this MCP server to an MCP client to analyze YouTube videos. You can generate a complete report that includes a summary, topics, entities, and even comments, or you can extract just the transcript summary. Use the commands below to create a report or transcript, and note that wrapping URLs in quotes is recommended for shells where URLs contain special characters.

How to install

Prerequisites: you need Python and pip installed on your system.

pip install mcp-youtube-intelligence
pip install yt-dlp  # for subtitle extraction

Additional notes and usage patterns

Basic usage works without an external LLM. For high-quality summaries, configure an LLM provider and model as described in the configuration section. You can run the following commands to generate a full report or a transcript summary.

mcp-yt report "https://www.youtube.com/watch?v=LV6Juz0xcrY"  # full report (needs LLM)
mcp-yt transcript "https://www.youtube.com/watch?v=LV6Juz0xcrY"  # transcript summary
mcp-yt report LV6Juz0xcrY  # using video ID only

Configuration and runtime notes

Configure your MCP client to connect to the YouTube Intelligence server via stdio. The following example shows how to register the server with a Claude Desktop, Cursor, or OpenCode client using uvx to run the MCP server.

{
  "mcpServers": {
    "youtube": {
      "command": "uvx",
      "args": ["mcp-youtube-intelligence"],
      "env": {
        "MYI_LLM_PROVIDER": "ollama",
        "MYI_OLLAMA_MODEL": "qwen2.5:7b"
      }
    }
  }
}

Troubleshooting and tips

If you encounter shell issues, wrap URLs in quotes. Install subtitle tool if missing. If you see SQLite database locked, ensure only a single server instance is running.

Available tools

get_video

Fetches metadata and generates a concise summary of the YouTube video.

get_transcript

Retrieves subtitles and can return a summary, full transcript, or segmented chunks.

get_comments

Fetches viewer comments and performs sentiment analysis; can provide a summarized view.

monitor_channel

Monitors a YouTube channel via RSS for new videos and updates.

search_transcripts

Searches within stored transcripts for relevant phrases or topics.

extract_entities

Extracts named entities (in Korean/English and other languages) from transcript data.

segment_topics

Performs topic segmentation to identify major themes across the video transcript.

search_youtube

Performs YouTube searches by keyword and returns results.

get_playlist

Analyzes a YouTube playlist and summarizes its contents.