# mcp-youtube-transcript [Health: Active]

**Category:** 🌐 Social Media  
**Repository:** https://github.com/jkawamoto/mcp-youtube-transcript  
**GitHub Stars:** 501  
**Views:** 13  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/mcp-youtube-transcript

## Description
An MCP server retrieving transcripts of YouTube videos

## Tools
Capabilities this server exposes over MCP:

- **get_transcript** — Retrieves the transcript of a YouTube video.
- **get_timed_transcript** — Retrieves the transcript of a YouTube video with timestamps.
- **get_video_info** — Retrieves the video information.
- **get_available_languages** — Retrieves the available languages for the video.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `uvx` (confidence: high):

```json
"mcpServers": {
  "mcp-youtube-transcript": {
    "command": "uvx",
    "args": ["--from","git+https://github.com/jkawamoto/mcp-youtube-transcript","mcp-youtube-transcript"]
  }
}
```

## Documentation

## What mcp-youtube-transcript does

mcp-youtube-transcript connects MCP-compatible AI applications to YouTube video transcripts and related metadata. Its four tools accept a full YouTube video URL and let an agent retrieve transcript text, timestamped segments, available transcript languages, or video information such as the title and channel.

Transcript requests support a preferred language code, with English used by default. The language listing tool can help an agent choose among manual and automatically generated transcript options before requesting content.

Long-form content is handled through response chunking. Transcripts larger than the default 50,000-character limit are split into responses, and the server returns a cursor that can be supplied to retrieve the next chunk. This is useful for lectures, conferences, interviews, and podcast episodes that may exceed an LLM context window.

## How it works

The server uses the Model Context Protocol over standard stdio. An MCP client starts the process and invokes tools with a YouTube URL plus any supported options. For transcript tools, the optional `lang` parameter selects the requested language and `next_cursor` requests a later portion of a paginated response.

The default response limit is 50,000 characters. You can change it with `--response-limit`; a negative value such as `-1` disables pagination and requests the complete transcript in one response.

Requests can use residential proxy services or ordinary HTTP and HTTPS proxies. Supported proxy configuration includes Webshare credentials, a ScrapingAnt API token, and custom proxy endpoints. ScrapingAnt access requires a paid subscription, and residential proxy usage consumes its credits.

## Setup and configuration

Install Astral `uv` to use the `uvx` launcher, then add the server to an MCP client's configuration. A typical command runs the package directly from its Git repository:

```json
{
  "mcpServers": {
    "youtube-transcript": {
      "command": "uvx",
      "args": [
        "--from",
        "git+https://github.com/jkawamoto/mcp-youtube-transcript",
        "mcp-youtube-transcript"
      ]
    }
  }
}
```

The repository also provides a Docker image through Docker Hub. Claude Desktop, Cursor, Goose, and LM Studio have setup paths described for this server; other standard MCP clients can use the stdio configuration when they support command-based servers.

Optional proxy environment variables include `WEBSHARE_PROXY_USERNAME`, `WEBSHARE_PROXY_PASSWORD`, `SCRAPINGANT_API_TOKEN`, `HTTP_PROXY`, and `HTTPS_PROXY`. The equivalent command-line flags are available for these settings. Add `--response-limit` to the command arguments when changing transcript pagination behavior.

## Tools and capabilities

- `get_transcript` returns plain transcript text. It accepts `url`, optional `lang`, and optional `next_cursor`.
- `get_timed_transcript` returns transcript segments with start times and durations. It accepts the same parameters as `get_transcript`.
- `get_available_languages` lists transcript languages for a supplied YouTube URL.
- `get_video_info` retrieves video metadata for a supplied YouTube URL.
- Automatic chunking and cursor-based pagination support transcripts longer than the configured response limit.
- Proxy configuration supports Webshare, ScrapingAnt, and generic HTTP or HTTPS proxy endpoints.

## Limitations and notes

YouTube may block automated requests from data-center or cloud-provider IP addresses. Residential or rotating proxies may be needed to avoid request blocking or IP bans. ScrapingAnt requires a paid web-scraping subscription for YouTube access.

Transcript availability and language choices depend on the target YouTube video. The server needs a full YouTube video URL for every tool call. The default language preference is English unless a different language code is supplied.

_Full upstream README: https://allmcps.com/mcp/mcp-youtube-transcript/readme_

