How to Query Your YouTube Library from Cursor and Claude with MCP

You've watched hundreds of hours of YouTube tutorials. React patterns, system design talks, deployment guides, debugging walkthroughs. All that knowledge is locked inside videos you can't search.

Meanwhile, your AI coding assistant — Cursor, Claude Code, Claude Desktop — has no idea any of it exists. When you ask it a question, it draws from its training data. It can't reference that specific video where Theo explained exactly why your hydration error is happening.

Skip's MCP server fixes this. It connects your video knowledge base to your AI tools, so your coding assistant can search across everything you've watched and give you answers with timestamps.

What Is MCP and Why Should You Care?

MCP (Model Context Protocol) is an open standard created by Anthropic that lets AI assistants connect to external tools and data sources. Think of it like USB for AI — a standardized way to plug capabilities into your assistant.

Without MCP, your AI assistant is limited to its training data and whatever you paste into the chat. With MCP, it can reach into external systems — databases, APIs, file systems, and in this case, your personal video knowledge base.

Skip implements an MCP server that exposes six tools to your AI assistant:

• search_knowledge_base — Semantic search across all your saved videos, returning relevant segments with exact timestamps
• discover_creators — Find expert YouTube educators on any topic
• import_video — Add a YouTube video to your library without leaving the editor
• import_creator — Bulk import videos from a channel
• list_videos — Browse your saved videos with filters
• get_video_summary — Get an AI-generated summary of any saved video

The Problem: Your AI Assistant Has Amnesia About Your Learning

Here's a scenario every developer knows. You're debugging an issue. You know you watched a video that covered this exact problem — the fix involved something with middleware configuration. But which video? You watched it three weeks ago. The title was something about "Next.js production tips" or maybe "deployment gotchas."

Without MCP, your options are:

1. Scroll through your YouTube history trying to find it
2. Ask ChatGPT/Claude and hope their training data covers it
3. Google the error message and start from scratch
4. Rewatch candidate videos at 2x speed, scrubbing for the right section

With Skip's MCP server connected, you just ask your coding assistant: "Search my videos for middleware configuration issues in Next.js production." It searches your entire library and returns the exact timestamp where the answer lives.

Setup: Connecting Skip to Your Editor

The setup takes about two minutes. You need a Skip account and an API key.

Step 1: Get Your API Key

Sign in at getskip.dev, go to your API Keys page, and generate a key. It starts with sk_live_.

Step 2: Add the MCP Config

For Cursor

Create or edit ~/.cursor/mcp.json:

{
  "mcpServers": {
    "skip": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://api.getskip.dev/mcp/sse"],
      "env": {
        "SKIP_API_KEY": "sk_live_your_key_here"
      }
    }
  }
}

For Claude Code (CLI)

Add Skip to your project's .mcp.json or global MCP config:

{
  "mcpServers": {
    "skip": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://api.getskip.dev/mcp/sse"],
      "env": {
        "SKIP_API_KEY": "sk_live_your_key_here"
      }
    }
  }
}

For Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\\Claude\\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "skip": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://api.getskip.dev/mcp/sse"],
      "env": {
        "SKIP_API_KEY": "sk_live_your_key_here"
      }
    }
  }
}

Step 3: Restart and Verify

Restart your editor. Ask: "What Skip tools do you have access to?" You should see the six tools listed above. If not, check that your API key is valid and the config file path is correct.

Real Usage: Developer Workflows

Here's where it gets interesting. These aren't hypothetical — they're the workflows that make MCP integration worth the two-minute setup.

Debugging with Your Video Library

You hit a cryptic error. Instead of context-switching to Google, ask your assistant:

"Search my videos for [error message or concept]"

It searches semantically — not just keywords. So "why is my React component rendering twice" finds segments about StrictMode, even if the video title never mentions "rendering twice."

Referencing Tutorials While Coding

You're implementing a feature and remember a tutorial that showed a clean pattern for it:

"What did ThePrimeagen say about error handling in Rust?"

Your assistant searches your library and returns the relevant section with a timestamp. You get the expert's explanation without leaving your editor or rewatching the video.

Building a Knowledge Base on the Fly

You find a great tutorial while researching. Import it without switching windows:

"Import this video to my library: https://youtube.com/watch?v=dQw4w9WgXcQ"

Skip processes the video, generates embeddings, and makes it searchable. Next time you (or your AI assistant) need that information, it's there.

Discovering New Learning Resources

Working on something unfamiliar? Find the best educators:

"Find YouTube creators who teach Kubernetes networking"

Skip's discovery engine finds relevant channels that YouTube's algorithm might never surface for you. Import their best videos and instantly expand your searchable knowledge base.

Why This Matters for Developer Productivity

The average developer watches 5-10 hours of technical YouTube content per week. That's 250-500 hours per year of accumulated knowledge that's effectively inaccessible after you close the tab.

MCP integration changes the economics. Every video you've watched becomes a reference you can query mid-flow. Your coding assistant goes from "helpful but generic" to "helpful and informed by everything you've specifically studied."

The compound effect is real: the more videos in your library, the more valuable every query becomes. After a few weeks, you stop thinking of Skip as a tool you use — it's just part of how your AI assistant works.

Getting Started

The fastest path:

1. Sign up for Skip (free tier includes MCP access)
2. Import 5-10 of your most-referenced YouTube tutorials
3. Add the MCP config to Cursor or Claude
4. Next time you're debugging, ask your assistant to search your library instead of Googling

The free tier gives you 50 videos and 100 messages per month — enough to see whether this changes how you work. Pro ($10/mo) unlocks 500 videos and 2,000 messages if you go all in.

Stop rewatching. Start querying.