How to give ChatGPT or Claude access to a YouTube video via MCP

How Plugsky MCP works in practice

Everything in this guide runs on Plugsky’s live MCP endpoint (https://plugsky.com/mcp) or the npm bridge (npx -y @plugsky/mcp). A client request flows: JSON-RPC over Streamable-HTTP → auth (API key or OAuth 2.1 + Dynamic Client Registration with fine-grained scopes) → tool router → the service you asked for. Model calls run behind a 7-tier provider failover chain; browsing renders with headless Chromium; transcripts come from real captions and Groq Whisper; images generate on FLUX; code runs in a sandbox with no network; memory persists in your account (Postgres-backed); and every tool call is metered per user on your dashboard.

Relevant live tools for this article: plugsky_youtube_transcript (real captions) and plugsky_video_ingest.

How to give ChatGPT or Claude access to a YouTube video via MCP

Direct answer

To let ChatGPT or Claude read a YouTube video in 2026, install a YouTube MCP server (youtube-transcript.ai, sinco-lab/mcp-youtube-transcript, or Plugsky's plugsky_youtube_transcript). Paste the video URL in chat; the MCP fetches the auto-captions (or runs Whisper if none exist) and returns the transcript for the LLM to read. Total time: 30 seconds to install, seconds per video.

Key facts

What MCPs do this	youtube-transcript.ai, sinco-lab MCP, Plugsky plugsky_youtube_transcript
Do they use YouTube captions?	Yes when present
What if no captions?	Whisper transcription fallback
Free?	Yes (youtube-transcript.ai + Plugsky free tier)
Supported clients	Claude, ChatGPT, Cursor, VS Code, Windsurf, Zed
Max video length	Practically 4+ hours; longer needs chunking
Rate limits	YouTube caption fetches ~free; Whisper fallback is minute-metered
Language support	60+ (whatever YouTube captions or Whisper support)
Bonus: chapter extraction	Yes on Plugsky (plugsky_video_transcript)

TL;DR

• Fastest path: add youtube-transcript.ai as a remote MCP in Claude Desktop — zero setup.
• Best all-in-one: use https://plugsky.com/mcp with plugsky_youtube_transcript — also gives you 36 models, RAG, and 143 more tools.
• Self-host option: sinco-lab/mcp-youtube-transcript for full local control.
• Once installed, users paste the YouTube URL and ask "summarize this video" — the LLM invokes the tool, gets the transcript, and answers.

How it works

Option 1: Plugsky plugsky_youtube_transcript

1. Add https://plugsky.com/mcp to your client (see [add MCP to Claude](/blog/add-mcp-to-claude)).
2. In chat: "Use plugsky_youtube_transcript on https://youtube.com/watch?v=xyz and give me a 5-bullet summary."
3. Response includes transcript + chapter breakdown + speaker diarization when detectable.

Advantages: one URL, 36 models to summarize with, RAG ingestion (send the transcript straight into a collection).

Option 2: youtube-transcript.ai (hosted, free, zero setup)

1. In Claude Desktop claude_desktop_config.json:

```json

{ "mcpServers": { "youtube": { "url": "https://mcp.youtube-transcript.ai" } } }

```

1. Restart Claude. Paste any YouTube URL and ask for a summary.

Advantages: zero auth, always-free.

Option 3: Self-host sinco-lab/mcp-youtube-transcript

{
  "mcpServers": {
    "youtube-local": {
      "command": "npx",
      "args": ["-y", "@sinco-lab/mcp-youtube-transcript"]
    }
  }
}

Advantages: no external service, local privacy.

Common patterns once it's installed

• Instant summary — "Summarize this video in 5 bullets."
• Timestamped Q&A — "At what timestamp does the speaker discuss pricing?"
• Multi-video comparison — "Compare how these three podcasts define agentic AI." (Plugsky can call the tool multiple times per turn.)
• Ingest into RAG — "Add the transcript to my research-notes collection." (Plugsky-only.)
• Language translation — pull English captions, translate to Arabic with plugsky-arabic.

Comparison of the top three YouTube MCPs

Setup time	30s (Bearer key)	10s (zero auth)	60s (npx)
Whisper fallback if no captions	✅	✅	⚠️
Chapters + metadata	✅	⚠️	⚠️
RAG ingest one-shot	✅	❌	❌
Additional tools bundled	8 more	0	0
Cost	Free tier	Free	Free (self-host)
Privacy	Cloud	Cloud	Local

Common errors

"Video is private/unavailable"	The MCP can only fetch public videos
"Transcript in wrong language"	Force English with ?lang=en param on YouTube URL if using captions
"Truncation"	Very long videos (>4h) may need chunking; use plugsky_video_transcript with chunk_minutes=30
"Whisper too slow"	If server-side Whisper is timing out, prefer videos with existing captions

FAQ

Q: Do I need a YouTube API key?

A: For caption-fetch mode, no. For channel-wide or search operations you'd need Google's Data API, but that's beyond most MCPs' scope.

Q: Can Claude/ChatGPT play the video?

A: No — the MCP fetches the transcript, not the audio/video stream. The LLM reasons over the text.

Q: Does this work for Twitch/Vimeo/podcasts?

A: Plugsky's plugsky_video_transcript and plugsky_podcast_ingest work on any URL. Pure-YouTube MCPs are YouTube-only.

Q: What about age-restricted videos?

A: Fails on all pure-caption MCPs. Whisper-based fallback also fails because YouTube blocks the audio fetch.

Q: Is there rate limiting on YouTube?

A: Yes — YouTube throttles caption fetches. Hosted MCPs pool requests to avoid your IP getting blocked, which is why hosted is usually more reliable than self-host.

Q: Can I add YouTube ingest into a Plugsky RAG collection?

A: Yes — plugsky_url_ingest accepts a YouTube URL, transcribes it, and chunks it into any collection. One tool call, done.

Trust & sources

• Author: Mustafa Hasan.
• Last updated: {DATE}.
• References: youtube-transcript.ai, sinco-lab MCP, [Plugsky MCP](/mcp).
• Related: [Read any website via MCP](/blog/chatgpt-read-website-mcp) · [Best free MCP servers](/blog/best-free-mcp-servers-2026).