About youtube-summarize
An MCP server for retrieving YouTube transcripts and video information, with optional summarization instructions. It supports transcript formats, language preferences, playlist browsing, and filtered video search.
Key features
- 01Fetch transcripts in five formats
- 02Retrieve video metadata
- 03Add transcript timestamps
- 04Return reviewable summary instructions
- 05Browse playlists and search videos
- 06Select transcript languages
Use cases
- 01Read a video transcript in a chosen format
- 02Review summary instructions before LLM processing
- 03Find videos matching search filters
- 04Enumerate and sort playlist videos
MCP support
Exposes tools to MCP clients for transcripts, metadata, playlist listings, video search, and transcript-language discovery. Its summarization tool returns instructions and transcript content for the client LLM.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
Tools
Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.
Details
From the project READMEyoutube-summarize
MCP server that fetches YouTube video transcripts and optionally summarizes them.

Features
- Fetch transcripts in multiple formats (text, JSON, SRT, WebVTT, pretty-print)
- Video metadata — title, description, channel, upload date, duration, views, chapters (via yt-dlp)
- Optional timestamps in plain-text transcripts
- Summarize videos — returns transcript with the prompt clearly broken out for human review before the LLM acts on it
- List playlist videos — fast enumeration of any public playlist with optional sorting
- Search videos — clean ranked results with sort and filter options, without the ads, Shorts shelves, and personalized recommendations of the YouTube website
- List available languages for any video's transcripts
- Flexible URL parsing — accepts full YouTube URLs (
youtube.com/watch?v=,youtu.be/,youtube.com/embed/,youtube.com/shorts/) or bare video IDs - Multi-language support — request transcripts in specific languages with fallback priority
Tools
get_transcript
Fetch a YouTube video's transcript. By default the response is prefixed with a [METADATA] block (title, channel, published, duration, views, description); pass include_metadata=false for transcript-only output.
| Parameter | Type | Default | Description |
|---|---|---|---|
url | string | required | YouTube video URL or video ID |
languages | string[] | ["en"] | Preferred languages in priority order |
format | string | "text" | Output format: text, json, pretty, webvtt, srt |
preserve_formatting | boolean | false | Keep HTML formatting tags in the transcript |
include_timestamps | boolean | false | When true with format="text", prefix each line with [HH:MM:SS]. Ignored for other formats (they already include timestamps). |
include_metadata | boolean | true | Prepend a [METADATA] block before the transcript. Pass false for transcript-only output. |
summarize_transcript
Fetch a transcript and return it with summarization instructions. The response is structured into clearly-labeled sections ([INSTRUCTIONS], [PROMPT_SOURCE], [VIDEO], [METADATA], [TRANSCRIPT]) so a human can review the prompt before letting the LLM act on it.
| Parameter | Type | Default | Description |
|---|---|---|---|
url | string | required | YouTube video URL or video ID |
prompt | string | (default prompt) | Custom summarization instructions |
languages | string[] | ["en"] | Preferred languages in priority order |
include_timestamps | boolean | false | Prefix each transcript line with [HH:MM:SS]. |
include_metadata | boolean | true | Include a [VIDEO] block with title, channel, published, duration, views, and description. |
get_video_metadata
Fetch metadata (title, description, channel, upload date, duration, views, tags, chapters, etc.) for a YouTube video. Backed by yt-dlp.
| Parameter | Type | Default | Description |
|---|---|---|---|
url | string | required | YouTube video URL or video ID |
list_playlist_videos
List the videos in a YouTube playlist (title, ID, channel, duration, views, URL). Per-video metadata is intentionally lean so the call stays fast even for large playlists; call get_video_metadata with a specific video ID for full detail.
| Parameter | Type | Default | Description |
|---|---|---|---|
url | string | required | YouTube playlist URL (with ?list=...) or bare playlist ID |
limit | integer | 500 | Maximum videos to return |
sort_by | string | "index" | One of: index (playlist order), title, duration, views. upload_date is not supported. |
order | string | "asc" | asc or desc |
search_videos
Search YouTube for videos and return a clean ranked list (title, ID, channel, duration, views, URL) — only actual videos, with no ads, recommendation shelves, or personalization. Relevance ordering still comes from YouTube's backend. Pair with get_transcript or summarize_transcript on a result's ID or URL.
| Parameter | Type | Default | Description |
|---|---|---|---|
query | string | required | Search terms |
limit | integer | 10 | Maximum results to return |
sort_by | string | "relevance" | One of: relevance, date (newest first), views, rating |
uploaded | string | "any" | Filter by upload time: any, hour, today, week, month, year |
duration | string | "any" | Filter by length: any, short (<4 min), medium (4–20 min), long (>20 min) |
list_transcripts
List available transcript languages for a video.
| Parameter | Type | Default | Description |
|---|---|---|---|
url | string | required | YouTube video URL or video ID |
Installation
Quick start (recommended)
uvx youtube-summarize
Claude Desktop
Add to your claude_desktop_config.json:
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"youtube-summarize": {
"command": "uvx",
"args": ["youtube-summarize"]
}
}
}
Claude Code
claude mcp add youtube-summarize -- uvx youtube-summarize
Other MCP clients
Run the server over stdio:
uvx youtube-summarize
Prerequisites
- Python 3.13+
- uv package manager
Development
# Install dependencies
uv sync
# Launch the MCP inspector (web UI for testing tools)
uv run mcp dev main.py
License
MIT
mcp-name: io.github.zlatkoc/youtube-summarize
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.