io.github.yjhann33-design/korea-stock-insight-mcp

Korea Stock Insight MCP

mcp-name: io.github.yjhann33-design/korea-stock-insight-mcp

English-first MCP — built for global LLM analysts, not Korean fluency speakers.

Drop-in for Claude Desktop, Cursor, Cline, and any MCP-compatible client. Ask in English, get Korean filings, financials, and KOSPI/KOSDAQ prices in one tool call chain — DART corp_code lookup, disclosures, XBRL statements, daily OHLCV.

What you get

DART (전자공시시스템) is South Korea's electronic disclosure system — the SEC EDGAR equivalent, operated by the Financial Supervisory Service (FSS).

Why English-first matters

Korean public filings are written in Korean, ticker codes are 6-digit numerics with no global mnemonic, and the disclosure system (DART) has no English equivalent of EDGAR's full-text search. For a non-Korean analyst, that's three friction layers before any actual research.

This server collapses them. Tool descriptions, parameter names, and examples are all written for an LLM that doesn't read Korean — company resolution accepts English names ("Samsung Electronics", "Naver"), responses include English context where useful, and the workflow is one Claude/Cursor turn instead of three browser tabs.

• English-first tool descriptions — names, params, examples all in English
• DART + KRX in one place — disclosures, financials, daily prices through a single MCP
• One-line install, no account — uvx korea-stock-insight-mcp and you're done (only a free DART key needed)
• Local stdio, MIT, no telemetry — your queries don't leave your machine

Quick start (Claude Desktop)

1. Get a free DART API key

Register at opendart.fss.or.kr and grab your API key from the user page. Instant, free, no credit card.

2. Add to your Claude Desktop config

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

{
  "mcpServers": {
    "korea-stock-insight": {
      "command": "uvx",
      "args": ["korea-stock-insight-mcp"],
      "env": {
        "DART_API_KEY": "YOUR_DART_KEY_HERE"
      }
    }
  }
}

If you don't have uvx, install uv first (one-liner on macOS/Linux/Windows), or use pipx:

{
  "mcpServers": {
    "korea-stock-insight": {
      "command": "pipx",
      "args": ["run", "korea-stock-insight-mcp"],
      "env": {"DART_API_KEY": "YOUR_DART_KEY_HERE"}
    }
  }
}

3. Restart Claude Desktop and try

"What were Samsung Electronics' most recent disclosures? Pull their last quarter's revenue and operating profit."

Claude will chain get_corp_code("Samsung Electronics") → get_disclosure_list("005930") → get_financial_statement(corp_code, 2025, "FY") automatically.

Sample queries (30 seconds, copy-paste into Claude)

Three workflows global analysts actually run on Korean equities. Each one chains 2–4 tool calls automatically.

1. Memory-cycle peer comparison

"Compare Samsung Electronics and SK Hynix on revenue, operating margin, and capex for the most recent annual filing. Highlight where they diverge."

Resolves both names to corp_codes → pulls latest XBRL → returns a side-by-side. Useful for tracking the HBM/DRAM cycle.

2. Disclosure radar before a catalyst

"Show me Naver's last 30 days of DART filings. Flag anything about earnings, M&A, share buybacks, or executive changes."

get_disclosure_list with date range, then Claude triages titles. No Korean reading required — the tool description tells Claude what filing types to look for.

3. Price + filing context in one turn

"What did Kakao file most recently, and how has the stock moved on KOSPI over the past 90 days?"

Combines get_disclosure_list + get_stock_quote so you get the news and the price reaction in the same answer.

Other examples

• "Pull LG Energy Solution's last annual report and summarize the management discussion."
• "What's the KODEX 200 ETF's 6-month trend?" (works with ETFs too)
• "For Hyundai Motor, list every disclosure tagged as a major decision (주요사항보고서) in 2025."

Roadmap

• v0.2 — Full disclosure body extraction (zipped XBRL → plain text)
• v0.3 — Multi-company batch queries, sector aggregations
• Future — Optional hosted tier with English-summarized disclosures and webhook push (no ETA; track progress in GitHub Issues)

Feedback & contributions

This is solo-maintained, and stdio mode means I can't see who's using it. The only signals I get are PyPI downloads, GitHub stars, and what you tell me.

If this helps you, please let me know:

• 🐛 Open an issue — bugs, missing tools, broken queries
• 💬 Discussions — "anyone analyzed [ticker]?", "is this the right tool for X?", general questions
• ⭐ Star the repo — easiest way to signal demand and shape the roadmap

If you actually analyze Korean stocks with Claude/Cursor, I'd genuinely love to hear what queries you run and what's missing.

License

MIT. See LICENSE.

Acknowledgments

• DART OpenAPI by Korea's Financial Supervisory Service
• KRX daily prices via yfinance
• Model Context Protocol by Anthropic