YouTube MCP Server

The all-in-one MCP server for YouTube. Plug any AI agent — Claude, Cursor, ChatGPT, VS Code, or your own — into a single endpoint and pull live data from any YouTube video or channel: transcripts, metadata, thumbnails, full video and MP3 downloads, and channel contact emails.

What you get

• Full transcripts — text plus timestamped segments in any available language, manual or auto-generated, with language fallback and a list of every track the video offers
• Video metadata — title, description, uploader and channel info, duration, view/like/comment counts, upload date, language, age limit, live status, categories, tags, chapters, top thumbnails
• Thumbnail downloads — every available size (maxres, sd, hq, mq, default) saved as JPGs in the run's storage
• MP4 video downloads — pick 360p, 720p, 1080p, 1440p, or "best available", saved straight to the run's storage
• MP3 audio rips — 128, 192, or 320 kbps, saved straight to the run's storage
• Channel contact emails — scans the About page plus recent video descriptions, deduplicates, filters out no-reply/fake addresses, and reports the source for every email
• Bulk lookups — query up to 10 videos in a single transcript or metadata call, up to 20 thumbnails, up to 10 channels for emails
• One endpoint, six tools — no glue code, no per-tool integrations, no scraping infrastructure to maintain

Use cases

• AI agents and assistants that need YouTube data without building a custom scraper
• Podcast and content research — pull a transcript and metadata for any video in seconds
• Content reuse — turn a long video into a transcript, MP3, and thumbnail set ready for a blog post or social clip
• Archival and offline access — back up videos you have permission to use as MP4 or MP3
• Creator outreach and lead generation — pull contact emails from creators in a niche
• Competitor monitoring — track view counts, descriptions, and chapters across a list of videos

Tools exposed

Tool	Returns
get-transcript	Transcript text + timestamped segments for up to 10 videos per call
get-metadata	Title, description, uploader, channel, duration, counts, chapters, thumbnails, tags
get-thumbnail	JPG thumbnails saved to the run's Key-Value Store (up to 20 videos, 5 sizes each)
download-video	A single MP4 saved to the run's Key-Value Store (single video per call, up to 2 GB)
download-mp3	A single MP3 saved to the run's Key-Value Store (single video per call, up to 500 MB)
get-email	Contact emails extracted from channel + recent video descriptions, up to 10 channels per call

All saved files (thumbnails, MP4, MP3) come back as fileUrl fields in the response — fetch them with the same Apify API token you used to call the MCP server.

How to use

Use Apify's hosted MCP gateway — it generates ready-to-paste config snippets for Claude Desktop, Cursor, VS Code, ChatGPT, and any other MCP-compatible client:

mcp.apify.com/?tools=seemuapps/youtube-mcp

Open the link, sign in, and copy the snippet for your client.

Where to get your Apify API token

Go to console.apify.com/settings/integrations and copy your Personal API token. (Sign in or create a free account if you haven't already.)

The token authorizes calls to the MCP server, lets you fetch saved thumbnails / MP4s / MP3s from the run's Key-Value Store, and is what Apify uses to bill you for tool calls.

Pricing (pay-per-event)

Tool	Event	Charged
get-transcript	transcript-extracted	Once per video where a transcript was successfully returned
get-metadata	metadata-extracted	Once per video where metadata was successfully returned
get-thumbnail	thumbnails-fetched	Once per video where at least one thumbnail size was saved
download-video	video-downloaded	Once per MP4 successfully saved
download-mp3	audio-converted	Once per MP3 successfully saved
get-email	channel-scanned	Once per channel scanned
get-email	email-found	Once per unique email returned

Failed calls are not charged. Partial failures (e.g. one bad URL in a batch) are reported per-item without aborting the whole call.

File size and download notes

• Thumbnails are tiny (a few KB to ~150 KB) and stored as JPGs.
• Video downloads are capped at 2 GB per file and 1 hour per video by default. Override maxDurationSeconds (set 0 to disable) for longer content. Higher resolutions and longer videos take longer — a 1440p hour-long video can take several minutes to fetch.
• MP3 downloads are capped at 500 MB per file and 1 hour per video by default. A typical 10-minute video at 192 kbps is ~14 MB.
• All saved files are stored in the run's default Key-Value Store. Each tool response includes a fileUrl you fetch with your Apify API token to retrieve the bytes.

Notes

• Only public YouTube videos and channels are supported.
• Auto-generated captions are used when manual captions are not available; toggle preferManualCaptions on get-transcript to invert this.
• The default proxy mode is residential, which works for most YouTube content. Switch to auto if you want to try without a proxy first.