mcp-youtube-transcript MCP — Use Cases, Install & Live Demo

Why use it

Key features

4 tools: get_transcript, get_timed_transcript, get_video_info, get_available_languages
Auto-pagination on transcripts ≥50k chars (configurable via --response-limit)
Proxy support (Webshare, custom HTTP/HTTPS) for IP blocking workarounds
Language selection — fall back to auto-generated if no manual captions
Runs via uvx mcp-youtube-transcript or .mcpb bundle for Claude Desktop

Live Demo

What it looks like in practice

youtube-transcript.replay ▶ ready

0/0

Install

Pick your client

~/Library/Application Support/Claude/claude_desktop_config.json · Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "youtube-transcript": {
      "command": "uvx",
      "args": [
        "mcp-youtube-transcript"
      ],
      "_inferred": true
    }
  }
}

Open Claude Desktop → Settings → Developer → Edit Config. Restart after saving.

~/.cursor/mcp.json · .cursor/mcp.json

{
  "mcpServers": {
    "youtube-transcript": {
      "command": "uvx",
      "args": [
        "mcp-youtube-transcript"
      ],
      "_inferred": true
    }
  }
}

Cursor uses the same mcpServers schema as Claude Desktop. Project config wins over global.

VS Code → Cline → MCP Servers → Edit

{
  "mcpServers": {
    "youtube-transcript": {
      "command": "uvx",
      "args": [
        "mcp-youtube-transcript"
      ],
      "_inferred": true
    }
  }
}

Click the MCP Servers icon in the Cline sidebar, then "Edit Configuration".

~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "youtube-transcript": {
      "command": "uvx",
      "args": [
        "mcp-youtube-transcript"
      ],
      "_inferred": true
    }
  }
}

Same shape as Claude Desktop. Restart Windsurf to pick up changes.

~/.continue/config.json

{
  "mcpServers": [
    {
      "name": "youtube-transcript",
      "command": "uvx",
      "args": [
        "mcp-youtube-transcript"
      ]
    }
  ]
}

Continue uses an array of server objects rather than a map.

~/.config/zed/settings.json

{
  "context_servers": {
    "youtube-transcript": {
      "command": {
        "path": "uvx",
        "args": [
          "mcp-youtube-transcript"
        ]
      }
    }
  }
}

Add to context_servers. Zed hot-reloads on save.

claude mcp add youtube-transcript -- uvx mcp-youtube-transcript

One-liner. Verify with claude mcp list. Remove with claude mcp remove.

Use Cases

Real-world ways to use mcp-youtube-transcript

Turn a 60-minute YouTube talk into a 10-bullet summary with timestamps

👤 Researchers, content marketers, students ⏱ ~10 min beginner

When to use: You can't watch the full video but need the key points (and to cite them with timestamps).

Flow

Fetch timed transcript

Get timed transcript for https://www.youtube.com/watch?v=VIDEO_ID.✓ Copied

→ Transcript with [MM:SS] markers
Summarize with citations

Give me 10 bullets summarizing the main argument. Each bullet should end with a [MM:SS] pointer to where it's made in the video.✓ Copied

→ Bulleted summary with clickable-style timestamps

Outcome: Digestible summary you can share with a team Slack in 2 minutes.

Pitfalls

Auto-generated captions have errors on technical terms — For accuracy-critical quotes, verify against the actual video at the timestamp

Combine with: notion

Index a YouTube podcast's back catalog for semantic search

👤 Podcast fans, researchers following a specific show ⏱ ~45 min intermediate

When to use: You want to ask 'which episode did they talk about X?' across 100+ episodes.

Flow

List episode URLs

Here's the playlist URL. Get the 50 most recent video IDs.✓ Copied

→ Video ID list
Fetch + index

For each video, get_transcript and get_video_info (title, date). Ingest into local-rag as one doc per episode.✓ Copied

→ Indexed corpus
Query

In which episodes did the hosts discuss 'ring attention'? Give episode number and timestamp.✓ Copied

→ Hits with episode metadata

Outcome: A searchable podcast library you own and control.

Pitfalls

Some videos have no captions — get_available_languages first; skip if empty

Combine with: local-rag

Translate a talk from one language to another

👤 Non-English speakers, content localizers ⏱ ~10 min beginner

When to use: You want to read a Spanish talk in English without running ML translation yourself.

Flow

Grab original language transcript

get_available_languages for this URL. Use Spanish if available.✓ Copied

→ Spanish transcript
Translate

Translate this transcript to English, preserving paragraph breaks.✓ Copied

→ English text

Outcome: Read-worthy translated transcript.

Pitfalls

Tonal / idiomatic content translates awkwardly — For commercial localization, have a human pass

Combinations

Pair with other MCPs for X10 leverage

youtube-transcript + local-rag

Ingest a channel's transcripts into a searchable local index

For each video in this playlist, get transcript + metadata, then ingest_data into local-rag with source URL.✓ Copied

youtube-transcript + notion

Publish video summaries to a Notion knowledge DB

Summarize this talk and create a Notion page in 'Talks' with title, channel, date, and summary.✓ Copied

Tools

What this MCP exposes

Tool	Inputs	When to call	Cost
get_transcript	url_or_id: str, language?: str	You want text only, no timestamps	free
get_timed_transcript	url_or_id, language?	You need timestamps for citations or chapter generation	free
get_video_info	url_or_id	Metadata for context / indexing	free
get_available_languages	url_or_id	Check caption availability before attempting fetch	free

Cost & Limits

What this costs to run

API quota: YouTube doesn't officially expose transcripts — expect IP rate limits
Tokens per call: Transcripts 1k-50k tokens depending on length; paginated at 50k by default
Monetary: Free; proxies (Webshare) cost a few $/mo if you need them
Tip: Transcripts are often huge. Truncate via --response-limit or just ask for a specific time range.

Security

Permissions, secrets, blast radius

Credential storage: Proxy credentials if you use Webshare — env var

Data egress: Outbound to youtube.com (possibly via proxy)

YouTube actively blocks scraping. From datacenter IPs expect 429s — use a residential proxy (Webshare) if persistent.
Respect each video's copyright — transcripts are useful for summarization and research, not wholesale reproduction.

Troubleshooting

Common errors and fixes

HTTP 429 Too Many Requests

Your IP is rate-limited by YouTube. Configure Webshare proxy via env vars or wait 30-60 minutes.

Verify: Try the same URL from another IP

No transcript available

Video has no captions. Call get_available_languages first; if empty, skip.

Age-restricted / region-locked video

Not supported by this MCP. No workaround — use official YouTube Data API with auth.

Python install issues (uvx)

Install uv first: curl -LsSf https://astral.sh/uv/install.sh | sh. Requires Python 3.10+.

Verify: uv --version

Alternatives

mcp-youtube-transcript vs others

Alternative	When to use it instead	Tradeoff
YouTube Data API + official MCP wrapper	You need official, authenticated access	Requires Google API key; quota limits; doesn't directly give transcript text
yt-dlp MCP	You also want video/audio download, not just transcripts	Heavier; downloads media files

More

Resources

📖 Read the official README on GitHub

🐙 Browse open issues

🔍 Browse all 400+ MCP servers and Skills