Can Claude summarize YouTube videos longer than 2 hours?

Yes. YouTubeAPI.io efficiently returns transcripts of any length, and modern LLM context windows (Claude 3.5 Sonnet has 200k tokens, Gemini 1.5 Pro has 2M tokens) can easily ingest 4+ hour transcripts in a single prompt.

Does this MCP server work with Cursor IDE?

Yes! In Cursor, open Cursor Settings -> Features -> MCP Servers -> Add New MCP Server. Select command type and enter: npx -y @youtubeapi/mcp-server with your YOUTUBEAPI_KEY environment variable.

Model Context Protocol · 20265 min read·Updated 2026-11-05

YouTube MCP Server for Claude Desktop & Cursor IDE: Complete Setup Guide (2026)

Empower your LLM agents with real-time YouTube intelligence. Learn how to configure @youtubeapi/mcp-server in 60 seconds with Claude Desktop, Cursor, and Windsurf.

Get Free API Key (2,000 Free Credits) →Run Live Code Example ↓

$0.20 / 1K Videos · $0.10 / 1K Comments & Followers · No OAuth Needed

Instant Quickstart · Copy & Run in 30 Seconds

Claude Desktop MCP Configuration

Execute in Live Playground →

# Python equivalent of how the MCP agent processes queries:
import requests
res = requests.get("https://api.youtubeapi.io/v1/youtube/transcript",
                   params={"video_id": "dQw4w9WgXcQ"},
                   headers={"x-api-key": "yt_live_your_key"})
print(res.json()["data"]["full_transcript_text"][:200])

Key Engineering Takeaways (TL;DR)

✓The Model Context Protocol (MCP) by Anthropic establishes an open standard for connecting AI clients (Claude Desktop, Cursor, Windsurf) to live external data sources.
✓The official @youtubeapi/mcp-server package provides 6 pre-built tools: youtube_transcript, youtube_video_info, youtube_search, youtube_comments, youtube_channel, and youtube_channel_videos.
✓Allows AI models to summarize 3-hour podcasts, analyze user comment sentiment, and track competitor video feeds directly inside your desktop chat interface.
✓Uses standard YouTubeAPI.io keys with 2,000 free signup credits and zero recurring subscription mandates.

1. Why AI Agents Need a YouTube MCP Server

Until now, asking Claude Desktop or Cursor to analyze a YouTube video meant either pasting truncated manual transcripts or relying on hallucinated knowledge cutoffs.

With the Model Context Protocol (MCP), LLMs gain native 'function calling' capabilities on your local desktop. When you ask: 'Summarize the architectural takeaways from this 2-hour conference talk: https://youtube.com/watch?v=...', Claude automatically calls the youtube_transcript tool, reads the entire timed transcript, and delivers an evidence-backed summary with exact video timestamp citations.

2. 60-Second Setup in Claude Desktop & Cursor

Adding YouTubeAPI to Claude Desktop requires adding a single JSON snippet to your claude_desktop_config.json file.

Claude Desktop MCP Configuration

Location on macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

// Test the underlying endpoint directly:
curl -s -X GET "https://api.youtubeapi.io/v1/youtube/transcript?video_id=dQw4w9WgXcQ" \
  -H "x-api-key: yt_live_your_key"

Interactive Engineering Workbench · MCP Protocol

Verify real-time latency, pagination math & JSON schema

Videos scraped per day (2 Credits / video, up to 33/call)5,000 videos/day

Comments extracted per day (1 Credit / comment, up to 50/call)10,000 comments/day

Followers mapped per day (1 Credit / user, up to 200/call)20,000 followers/day

Our HTTP Calls/Day

452

Large page sizes

Legacy Provider Calls

1,334

30-item small pages

Official Research API

Within 1K Cap

1,000 req/day limit

Estimated Monthly Pipeline Cost (30 Days)

$72.73/ mo with Business bonus ($120.00 base rate)

Daily consumption: 40,000 Credits ($4.00/day)

YouTubeAPI.io (with volume bonus):$72.73 / mo

Legacy Scrapers ($1.50 / 1K calls):$60.03 / mo

Failed / 404 request fee:$0.00 (0 Credits)

Start Free (2,000 Credits) →Full Rate Card

Frequently Asked Questions

Can Claude summarize YouTube videos longer than 2 hours?: Yes. YouTubeAPI.io efficiently returns transcripts of any length, and modern LLM context windows (Claude 3.5 Sonnet has 200k tokens, Gemini 1.5 Pro has 2M tokens) can easily ingest 4+ hour transcripts in a single prompt.
Does this MCP server work with Cursor IDE?: Yes! In Cursor, open Cursor Settings -> Features -> MCP Servers -> Add New MCP Server. Select command type and enter: npx -y @youtubeapi/mcp-server with your YOUTUBEAPI_KEY environment variable.

More engineering guides

View all guides →

AI & RAG Architecture · 2026

YouTube Transcript API for RAG Pipelines & AI Agents: Complete 2026 Engineering Guide

7 min read →

Quota & Cost Engineering · Guide

Bypassing the YouTube Data API v3 10,000-Unit Daily Quota Limit in Production

6 min read →

Audience Intelligence · Tutorial

Extracting YouTube Video Comments & Audience Sentiment for AI Analysis

6 min read →

Python & Cloud Fix · 2026

How to Fix 'IpBlocked: YouTube is blocking requests from your IP' in Python (2026)

6 min read →