kordoc
A CLI and MCP server that converts Korean government and office documents (HWP, HWPX, PDF, Office formats and scanned images) into Markdown, with table reconstruction, form filling and document diffing for AI agents to read and act on.
kordoc parses HWP 3.x/5.x, HWPX, HWPML, PDF, XLS/XLSX, DOCX and PNG/JPG/WebP images (with automatic OCR) into Markdown, aiming to put Korean official documents into a form an LLM can read and analyze. It reconstructs complex merged and nested tables, reporting near-lossless accuracy against an independent extractor on a large HWPX corpus, and can detect and restore borderless tables in PDFs, including redline-style comparison tables in draft legislation.
It generates side-by-side comparisons showing what changed between two documents, including between HWP and HWPX versions, and can write AI-authored Markdown content back into an official document template.
It installs as an MCP server for AI clients such as Claude Desktop, Cursor, Claude Code, Windsurf, VS Code, Gemini CLI, Zed, Antigravity or Codex through an interactive setup wizard, as a Claude Code plugin skill, or used standalone as a CLI via npx, and requires only Node.js 20 or later.