Scrape and analyze videos from YouTube’s 2B+ monthly users, complete with rich metadata, statistics, and comments in one place.
What is YouTube MCP?
When you create a YouTube node with AI, Gumloop builds a custom node that already understands Apify’s YouTube Scraper. Describe the result you need in plain English and the node maps your prompt to the right API call, returns clean structured data, and is ready to reuse in any workflow.What Can It Do for You?
- Discover fresh, high-performing videos for competitive research or content inspiration
- Enrich any video URL with views, likes, duration, tags, and more
- Pull an entire channel or playlist into Google Sheets for batch analysis
- Capture real user comments for sentiment analysis, product feedback, or testimonial mining
Available Tools
| Tool | What It Does | Example Use |
|---|---|---|
| Search Videos | Search YouTube by keyword with filters like upload date, type, and video length. Returns title, URL, views, and more. | ”Search videos about ‘Python pandas tutorial’ uploaded after 2024-01-01, return title, url, view_count, publish_date” |
| Get Video Details | Fetch full metadata for a specific video by URL or ID, including statistics and content details. | ”Get details for https://youtube.com/watch?v=dQw4w9WgXcQ, return title, channel_name, view_count, like_count, published_at” |
| Get Channel Videos | List videos from a channel with complete metadata. Supports sorting and limits. | ”Get videos from channel ‘@MrBeast’, limit 10, sort by most_popular, return title, url, view_count” |
| Get Playlist Videos | Retrieve videos in a playlist with video-level data and playlist context. | ”Get videos from playlist ‘Learn React - Full Course’ by freeCodeCamp, limit 20, return title, url, duration” |
| Get Channel Details | Return channel-level stats such as subscriber count, total videos, description, and creation date. | ”Get channel details for ‘@mkbhd’, return channel_name, subscriber_count, total_videos, description” |
| Get Video Comments | Collect comments from a video with author info, likes, and reply counts. | ”Get top 30 comments for video titled ‘iPhone 15 Review’ by MKBHD, return author, comment_text, like_count, published_time” |
Credit Costs
| Tool | Credits Per Use |
|---|---|
| Search Videos | 3 credits per item |
| Get Video Details | 4 credits per item |
| Get Channel Videos | 3 credits per item |
| Get Playlist Videos | 3 credits per item |
| Get Channel Details | 5 credits per item |
| Get Video Comments | 5 credits per item |
How to Use
1
Create Your YouTube MCP Node
Go to your node library, search for YouTube, and click “Create a node with AI”
2
Add Your Prompt
Drag the YouTube node onto the canvas and describe the single action you want it to perform.
3
Test Your Node
Run the node. Review the output to confirm it contains the fields you requested.
4
Save and Reuse
Save the working node to your library and drop it into any future workflow.
Example Prompts
Here are some prompts that work well with YouTube MCP: Search for Trend AnalysisStart with a narrow filter and small limit (for example, “limit 10”) to preview the structure, then increase the limit once you are happy with the fields.
Troubleshooting
If your YouTube MCP node isn’t behaving as expected, try these best practices:Keep Prompts Simple and Specific
- Good: “Search videos about ‘Stock market analysis’ uploaded after 2025-01-01, return title, url”
- Bad: “Search for the top stock market videos, and summarize each video”
While the complex prompt may work, it is more efficient to break it into separate nodes. YouTube MCP works best with focused, single-action prompts.
Match What YouTube Can Do
- Good: “Get video comments for
Video URL, limit 20, return author, comment_text” - Bad: “Download the video, transcribe it, and translate the transcript to French”
YouTube MCP excels at data retrieval. For transcription or translation, connect it to an appropriate node like an Ask AI node in your workflow.
Break Complex Tasks Into Steps
Instead of trying to do everything in one prompt (which can cause timeouts and errors):1
Step 1: Search Videos
Search videos about electric cars uploaded in August 2025, limit 50, return title, url, view_count
2
Step 2: Fetch Transcript
Given a video URL fetch the video transcript
3
Step 3: Categorize Sentiment using the AI Categorizer node
Classify each transcript as positive, neutral, or negative (Categorizer)
4
Step 4: Share to Slack using the 'Slack Message Sender' node
Send positive summaries to the #ev-insights Slack channel (Slack Message Sender)
In your workflow, connect these nodes sequentially. The video URLs from Step 1 feed into Step 2, sentiment labels from Step 3 determine which summaries are sent in Step 4.
Focus on Data Retrieval
YouTube MCP is great at getting information from YouTube. For analysis or content creation, pair it with other nodes. Example:- Good prompt: “Get playlist videos for PL12345, return title, duration”
- Bad prompt: “Get playlist videos, write a blog post summarizing each one, and publish it to WordPress”
Use Ask AI for writing and a Call API or Run Code node to publish content. Keep the YouTube node focused on data extraction.
Troubleshooting Node Creation
Empty Outputs
Empty Outputs
In the node creation window, click “Request changes” and ask the AI to add debug logs and verify the API response.
Incorrect Results
Incorrect Results
In the node creation window, click “Request changes” and describe what you expected versus what you received.
Errors
Errors
First click “Fix with Gummie”. If multiple attempts fail, simplify your prompt or contact support.
Iterate with Request Changes
Iterate with Request Changes
MCP node creation often requires a few tweaks. Use “Request changes” (in the node creation window) to refine filters, output fields, or pagination.
Need More Help?
- Watch What are MCP Nodes video tutorial
- Check out MCP Best Practices in Gumloop University
- Join the Gumloop Community for support
- View the YouTube MCP setup guide for Claude and Cursor
- Contact support at support@gumloop.com