YouTube Transcript MCP Server

quicktube-mcp is a Model Context Protocol server that lets AI agents read YouTube videos. Add it to Claude Code, Claude Desktop or Cursor and your agent can fetch a transcript, summarize a video, find its best moments, write a quiz or fact-check it, all from a prompt.

Get an API key See every MCP tool

Set it up

1
Create an API key
Sign up (new accounts get 400 free credits once they verify their email) and create a key in Settings → API Keys. New keys can fetch transcripts and run the AI tools.
2
Add the server to your client
One command for Claude Code, or a JSON snippet for Claude Desktop and Cursor. Both run the package with uvx, so there is nothing else to install.
3
Ask your agent about a video
"Get the transcript of this video and list every tool the speaker mentions." The agent picks the right QuickTube tool itself.

What your agent can do

get_transcript

The full transcript text with its language and whether it was auto-generated. 1 credit.

summarize_video and get_best_sections

An AI summary, or the most valuable sections with timestamps. 5 credits each.

generate_quiz and fact_check_video

Multiple-choice questions with explanations, or a verdict on each of the video’s factual claims. 5 credits each.

chat_with_video and more

Ask questions about a video (1 credit each), detect speakers, search YouTube and read comments. Tick “Also allow saved sessions and credit balance” when you create the key to let the agent check your balance too.

Claude Code

Run this once in a terminal. Add --scope user to make the server available in every project instead of just the current one.

Terminal

claude mcp add quicktube \
  --env QUICKTUBE_API_KEY=qt_your_api_key_here \
  -- uvx quicktube-mcp

Claude Desktop and Cursor

Both read the same mcpServers JSON. In Claude Desktop, open Settings → Developer → Edit Config to find claude_desktop_config.json, add the block below, and restart the app. In Cursor, put it in ~/.cursor/mcp.json for all projects or .cursor/mcp.json in one project.

claude_desktop_config.json or mcp.json

{
  "mcpServers": {
    "quicktube": {
      "command": "uvx",
      "args": ["quicktube-mcp"],
      "env": {
        "QUICKTUBE_API_KEY": "qt_your_api_key_here"
      }
    }
  }
}

Prefer pip?

uvx downloads and runs the package on demand, and needs uv installed. If you would rather install it into a Python 3.10+ environment yourself:

Terminal

pip install quicktube-mcp
# then use "command": "quicktube-mcp" with no args in the JSON config

Why give an agent YouTube transcripts

Language models can’t watch videos, and most can’t fetch YouTube captions on their own. With the transcript as text the agent can answer questions about a talk, pull the commands out of a coding tutorial, compare what two videos say, or turn a conference session into notes in your repo, without you copying and pasting.

Every tool call is billed to your QuickTube credits at the same price as the REST API, and the server only acts with the scopes you gave its key.

Frequently asked questions

Which AI tools can use the QuickTube MCP server?

Any client that supports local (stdio) MCP servers, including Claude Code, Claude Desktop and Cursor. The server is a Python package, quicktube-mcp, that you run with "uvx quicktube-mcp" or install with "pip install quicktube-mcp". It authenticates with your QuickTube API key.

How much does the YouTube transcript API cost?

Each transcript fetched with an API key costs 1 credit, including transcripts served from cache. AI endpoints cost the same as in the web app, for example 5 credits for a summary. A free account comes with 400 credits once your email is verified, and more are sold as one-time packs with no subscription.

Can AI agents use QuickTube?

Yes. QuickTube has a REST API and an MCP (Model Context Protocol) server, installable with "pip install quicktube-mcp", that let tools such as Claude Code and Claude Desktop fetch transcripts, summarize videos and generate quizzes. Fetching a transcript through the API costs 1 credit.

What if a video has no captions?

QuickTube reads the captions YouTube already has, either uploaded by the creator or generated automatically. It does not transcribe audio itself, so a video with captions turned off can't be transcribed or analyzed.