YouTube MCP Server

Connect Claude to your YouTube channel. Read analytics, fetch full video metadata (including unlisted/private/draft), search your uploads, and update titles/descriptions/tags — all from Claude Desktop, Claude Code, or any MCP client.

What Is This?

A Model Context Protocol (MCP) server that connects Claude to the YouTube Data API v3 and YouTube Analytics API via OAuth2.

Instead of copying data from YouTube Studio into AI tools, just ask Claude:

• "Pull the full title, description and tags for this video"
• "What are my top performing videos this month?"
• "Update the title and tags on my latest upload"
• "Where is my traffic coming from?"

Claude reads your real private channel data and can update video SEO directly.

Claude  →  YouTube MCP Server  →  YouTube APIs  →  Your Channel Data
(you)        (this repo)           (OAuth2)         (stays local)

Tools Available (10 total)

Video Metadata (Read + Write)

Channel Analytics

Install

Option A: npx (Zero Install — just run it)

npx youtube-studio-mcp

Option B: Global Install

npm install -g youtube-studio-mcp

Connect to Claude

Claude Desktop

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

{
  "mcpServers": {
    "youtube-analytics": {
      "command": "node",
      "args": ["/full/path/to/youtube-studio-mcp/server.js"]
    }
  }
}

Or if installed via npm:

{
  "mcpServers": {
    "youtube-analytics": {
      "command": "npx",
      "args": ["-y", "youtube-studio-mcp"]
    }
  }
}

Claude Code (Terminal)

claude mcp add youtube-analytics node /full/path/to/youtube-studio-mcp/server.js

VS Code

Add to .vscode/settings.json:

{
  "mcp.servers": {
    "youtube-analytics": {
      "command": "npx",
      "args": ["-y", "youtube-studio-mcp"]
    }
  }
}

Restart Claude. Done!

Ready-to-Use Prompts

SEO Audit (any video)

Get the full details for this video: [paste URL or ID]
Check what keywords it's ranking for and suggest optimized title + tags

Channel Performance Report

Pull my channel overview, top 20 videos by watch time, 90-day analytics,
traffic sources and audience demographics. Give me a full performance report.

Video Topic Research

Get my top 20 videos by watch time. What patterns do you see —
topics, lengths, title styles? Suggest 10 new video ideas.

Update Video SEO

Search my videos for "elementor menu". Pull the full details.
Write an optimized title, description, and tags — then update it.

Audience Deep Dive

Show my full audience demographics — age, gender, countries, devices.
What content style and posting schedule fits my actual audience?

Underperformer Diagnosis

Get all my videos. Compare bottom 10 vs top 10 by views.
Why did the lower ones underperform? What would you change?

How the Skills + MCP Work Together

You: "Optimize this video: youtube.com/watch?v=abc123"

Claude:
  1. youtube-seo-optimizer skill activates
  2. Calls get_video_details → pulls current title, description, tags
  3. Checks YouTube SERP for current rankings
  4. Lists PROTECTED keywords (won't remove)
  5. Writes 3 new title options + full description + 20 tags
  6. Can call update_video_seo to apply changes directly

The skills tell Claude what to do. The MCP tools give Claude access to your data. Together they create a complete YouTube AI workflow.

Install Skills

Claude Desktop / Claude Code:

# Copy a skill to your Claude skills directory
cp skills/youtube-seo-optimizer.md ~/.claude/skills/

Or install all 8:

cp skills/youtube-*.md ~/.claude/skills/

Files

youtube-studio-mcp/
├── server.js          ← MCP server (10 tools)
├── auth.js            ← Run once to link YouTube account
├── package.json       ← Dependencies + npm config
├── skills/            ← 8 AI skills for YouTube creators
├── LICENSE            ← MIT
├── credentials.json   ← YOU add this (never commit!)
└── tokens.json        ← Auto-created after auth (never commit!)

License

MIT — free to use, modify, share.

Built By

Hakan Özdemir & Claude Code