CK

chrisryugj/kordoc

Developer tools

This page was translated automatically. Technical details may differ from the original. View the original

2.8K stars 505 forks Quality 77 Trend 82MIT

I'll parse them all — HWP· HWPX·PDF· Office documents as Markdown. CLI·MCP server with automatic form filling and old and new comparison | Convert Korean documents (HWP, HWPX, PDF, Office) to Markdown — CLI and MCP server with form filling and diff

Overview

A document hell that would be the second most prestigious in Korea. A public official who endured seven years there created it. HWP 3.x/5.x, HWPX, HWPML, PDF, XLS, XLSX, DOCX — parse, compar, analyze, and generate all documents flooded by government offices. Conversational wizard: 1. Select the AI client number in use (Claude Desktop / Cursor / Claude Code / Windsurf / VS Code / Gemini CLI / Zed / Antigravity — Installed items show [Detected]) 2. Automatic patch of configuration files → Restart the client Windows also automatically wraps cmd /c npx files. No need for manual JSON editing. Restarting activates 10 document tools (parse_document, parse_table, fill_form, patch_document, generate_document, etc.). : The default PowerShell security policy is the standard behavior that blocks unsigned .ps1 (kordoc does not matter). You can use one of the following. (Safest) Search for the Windows key → cmd → Enter → Remain as it is in the black window: Administrator PowerShell: Then restart PowerShell → npx -y kordoc setup remains as is. When you mention .hwp/.hwpx or request to create or fill in official documents, the kordoc skill is automatically activated (internally, you call npx -y kordoc@^3 CLI — no separate installation required). * : Instantly converts HWP3 (old version), HWP (5.x), HWPX, HWPML, PDF, XLS, XLSX, DOCX files into Markdown. AI (LLM) creates the best state for reading and analyzing documents. * : Analyzes the structure of unlined PDFs or complex merged HWP tables to restore them to accurate markdown tables. The comparison table of new and new texts in the legal amendment PDF is fully restored (v3.16.2).

Overview (machine translated)

I’ll parse them all.

The second most sorrowful document hell in Korea. A public official who endured seven years there created it.

HWP 3.x·5.x, HWPX, HWPML, PDF, XLS· XLSX, DOCX, PPTX, PNG·JPG· Converts WebP into Markdown and structured data. Libraries· Can be used as CLI and MCP servers.

English · Detailed usage · Benchmark · Change history

  • 📊 PDF Public Bench Comprehensive 0.963 — OpenDataLoader-bench 200 document. Ranked 1st in the public parser comparison (recorded on 2026-09-29). Measurement Conditions
  • 🇰🇷 100% Korean official document table structure — Visible table of 2,429 original HWPX documents All 10,461 have identical structures. 4.22.0 Verification

If you found it useful, please click GitHub ⭐. It helps others find this tool.

▶ Click to play it on YouTube.

New Video: Kordoc Open Bench No.1 Mark · 2026-09-27 · 1 minute 31 seconds

Installation

Node.js 20 or above · macOS / Linux / Windows

The CLI can be run with npx kordoc without installation. PDF· OCR dependencies are installed by default, and excluding --omit=optional limits this functionality.

Connect to AI Agents

Claude Desktop· Claude Code· Cursor· Register MCP on installed clients such as Codex. When you restart the client, you can use 17 tools such as parsing, comparison, creation, and rendering.

Claude Code plugin:

Manual MCP Registration and Installation Troubleshooting

Quick Start

CLI

--jobs is a file-level parallel conversion and increases memory usage. Parallel Conversion Guide

For reports, plans, revision ceremonies, and Seoul policies, the quotation (> …하고자 함) immediately after the title # serves as the summary box (for the reorganization format, when there is a cover (basically), below the title box on the first page of the main text — cover:false page first # section header, so refer to ※). For the purpose of the report, write one sentence (commas allowed) within 3 lines; if it exceeds or contains two or more sentences, a warning is issued. For work reports, the > ▪ … citation box is used as the performance summary box regardless of position.

JavaScript / TypeScript

This page was translated automatically. Technical details may differ from the original.

Install

npx -y kordoc setup
New to MCP? Start with the mcpskill.org guide →