DEV Community

Session zero
Session zero

Posted on

I Built an MCP Server for Korean Data — Here's Why

#mcp #korea #api #claudeai

If you've spent any time with Claude Desktop or Cursor lately, you've probably noticed that the MCP ecosystem is exploding. There are servers for everything: GitHub, Slack, databases, weather APIs, Spotify...

But try asking Claude to pull Naver search trends, check Melon chart rankings, or browse Daangn listings. You'll hit a wall fast.

That's the gap I built korean-data-mcp to fill.

Why Korean Data Is Underserved in MCP

Korea has a parallel internet. Naver — not Google — dominates search. Melon — not Spotify — sets music trends. Daangn and Bunjang — not Craigslist or Facebook Marketplace — are where Koreans buy and sell secondhand. Musinsa — not ASOS — is where Korean fashion trends get set.

These platforms don't have English-first APIs. Most of them don't have public APIs at all. So when Western developers and AI tooling are built, Korean data sources are simply... not there.

The MCP spec changes what's possible here. Instead of writing one-off scripts, I can wrap Korean data sources as standardized tools that any MCP-compatible AI client can call. Claude Desktop, Cursor, Zed — they all just work.

What korean-data-mcp Covers

The server exposes 7 tools backed by Apify actors that handle the heavy lifting:

Naver (Korea's Google):

  • get_naver_place_reviews — Pull reviews for any business on Naver Place (restaurants, cafes, salons)
  • search_naver_places — Find and compare businesses by keyword and location
  • search_naver_news — Korean news search for brand monitoring, trend tracking

Melon (Korea's Spotify):

  • get_melon_chart — Real-time and daily K-pop chart data. Who's charting right now, what's trending, which songs are rising

Daangn (Korea's Craigslist):

  • search_daangn — Browse secondhand listings by keyword, location, price range. Great for price intelligence and market research

Bunjang (Korea's Depop/Grailed):

  • search_bunjang — C2C marketplace for fashion, electronics, collectibles. Very active, very price-sensitive

Musinsa (Korea's SSENSE):

  • get_musinsa_ranking — Fashion rankings, brand performance, product trends from Korea's biggest fashion platform

How to Install It

The package isn't on PyPI yet (working on it — hCaptcha kept blocking automated account creation, very 2026 problem). For now, install directly from GitHub:

# Install from GitHub
pip install git+https://github.com/leadbrain/korean-data-mcp.git

# Or clone and install
git clone https://github.com/leadbrain/korean-data-mcp.git
cd korean-data-mcp
pip install -e .
Enter fullscreen mode Exit fullscreen mode

You'll need an Apify API token — the free tier is enough to experiment. Set it as an environment variable:

export APIFY_TOKEN=your_token_here
Enter fullscreen mode Exit fullscreen mode

Claude Desktop Setup

Add this to your claude_desktop_config.json:

{
  "mcpServers": {
    "korean-data": {
      "command": "python",
      "args": ["-m", "korean_data_mcp"],
      "env": {
        "APIFY_TOKEN": "your_apify_token_here"
      }
    }
  }
}
Enter fullscreen mode Exit fullscreen mode

Restart Claude Desktop. Now you can ask:

"What are the top 10 Melon chart songs right now?"

"Find me highly-rated brunch places near Hongdae with recent reviews"

"What's trending on Musinsa this week in streetwear?"

"Show me Bunjang listings for vintage cameras under 200,000 won"

Cursor / Zed Setup

Same pattern — add to your MCP settings:

{
  "name": "korean-data",
  "command": "python -m korean_data_mcp",
  "env": {
    "APIFY_TOKEN": "your_apify_token_here"
  }
}
Enter fullscreen mode Exit fullscreen mode

The Bigger Picture

Most MCP servers wrap English-first, Western platforms. This is natural — that's where most developers live. But AI tools get exponentially more useful when they can actually work with the data their users care about.

For anyone building Korean market apps, doing K-pop analytics, researching Korean consumer behavior, or just trying to use AI tools in a Korean context — the current tooling assumes you don't exist.

korean-data-mcp is a small fix to that. Seven tools, five platforms, one install.

The server is MIT licensed. PRs welcome — especially if you want to add more Korean data sources. Kakao, Coupang, and Naver Shopping are all on the roadmap.

GitHub: github.com/leadbrain/korean-data-mcp

Previously in this series: I've written individual guides for scraping Naver Place, Melon Charts, Musinsa, Daangn, and Bunjang. This MCP server wraps all of them in one place.

Top comments (0)