MCP YouTube Intelligence

An MCP server and CLI for YouTube: server side transcript summaries, video reports, channel monitoring and comment analysis.

  • 53 ★ on the repository
  • Last commit
MCP server

Extended access

Why this level

  • Transcripts and comments go to the chosen LLM provider when one is connected.
  • The server contacts YouTube and runs `yt-dlp` on your machine.
All reasons and checks

janghyuckyun/mcp-youtube-intelligence

Install

Manual install

uv pip install mcp-youtube-intelligence

First install command from the README, then `pip install yt-dlp` for transcripts.

Access and checks

How we review

Extended accessThe tool runs code or makes network calls and only reads external data.

  • Transcripts and comments go to the chosen LLM provider when one is connected.
  • The server contacts YouTube and runs `yt-dlp` on your machine.

This is third-party code. Review the repository files before installing.

What it does

MCP YouTube Intelligence processes YouTube transcripts on the server side and gives the model 200–500 tokens of summary instead of 2,000–30,000 tokens of raw text. Besides summaries the server returns a structured report with topics, entities and comments, segments videos by topic, analyzes comment sentiment, searches transcripts and YouTube and processes video lists in batches. A `mcp-yt` CLI is included, for example `mcp-yt report URL` or `mcp-yt transcript ID --mode full`. Channels can be monitored by RSS with summaries forwarded to Slack or Discord. Results are cached in SQLite or PostgreSQL. A basic summary works without an LLM, and OpenAI, Anthropic or Google can be connected for better quality.

Who it is for. Researchers, marketers and authors who go through many YouTube videos and want to save agent tokens.

Good fit when

  • A short breakdown of an hour long video with topics and key entities is needed.
  • A competitor channel is put on monitoring and summaries of new videos arrive.
  • Comment sentiment shows the reaction to a release.
  • A batch of videos on a research topic is processed.

Not a fit when

  • You need a verbatim full transcript without processing.
  • The video has no subtitles and speech recognition is not set up.
  • You cannot install `yt-dlp` or reach YouTube.

Example request

Build a report on this YouTube video: summary, topics, entities and comment sentiment.

MCP

Transport
stdio
Authentication
not required

Limitations

Transcript extraction needs `yt-dlp` on `PATH`. Without an LLM only a basic summary by key sentence selection works, while a full report and a quality summary need a configured provider: OpenAI, Anthropic or Google with your key, so transcript text goes to that provider. The main README is in Korean and the English version is `README.en.md`. YouTube use is governed by its terms, so keep that in mind for bulk processing.

How to disable. Remove the MCP entry from the client configuration, the package with `pip uninstall mcp-youtube-intelligence` and the local SQLite cache.

FAQ

How do I install the server?

Run `uv pip install mcp-youtube-intelligence` or `pip install mcp-youtube-intelligence`, plus `pip install yt-dlp` for transcript extraction.

How is it different from ordinary YouTube MCP servers?

It compresses transcripts on the server and gives the model 200–500 tokens, and also builds reports, monitors channels and analyzes comments.

Does it work without an LLM?

Yes, a basic summary picks key sentences. Set up OpenAI, Anthropic or Google for higher quality.

How do I get the full transcript?

Run `mcp-yt transcript URL --mode full`: the text is saved to a file. The `chunks` mode splits it into parts.

How do I monitor a channel?

Run `mcp-yt monitor CHANNEL_ID --interval 3600`: the server checks RSS and summarizes new videos.

Discussion

No questions or reviews yet.

Russian stack

An MCP server publishes vertical video to 8 platforms at once, including VK Clips and Telegram, with 9:16 normalization and analytics

MCP serverBroad access3Repository stars

21 commands and 10 tools connect an agent to your YouTube channel: analytics, SEO written back to YouTube and a WordPress post.

SkillBroad access83Repository stars
Russian stack

Seven independent MCP servers in one repository: Yandex Direct, Yandex Webmaster, VK, Telegram, Click.ru, Search Console and YouTube

MCP serverBroad access2Repository stars

Self-hosted MCP server for xiaohongshu: publish notes and videos, search, read feeds, comment, like and favorite from an agent

MCP serverBroad access17.1KRepository stars
Official

A translation MCP from the LibreTranslate team: the agent detects a text's language, translates it and lists supported languages with three tools

MCP serverExtended access7Repository stars

README in short

The README presents the server as a token saving tool: a comparison table, architecture, use cases, CLI commands and LLM provider setup.

README badge

Report a problem