Skip to main content
hc
Submit
mcpSource-backedReview first Safety ✓ Privacy ✓

Kordoc MCP Server

MCP server for parsing Korean HWP, HWPX, HWPML, PDF, XLSX, and DOCX files into Markdown, with tools for metadata extraction, page ranges, tables, document diffs, form extraction, and HWPX form filling.

by chrisryugj·added 2026-06-06·
Claude CodeClaude Desktop
HarnessClaude CodeClaude Desktop
Review first — review before installing

Open the source and read safety notes before installing.

Safety notes

  • Kordoc MCP Server reads local document files by absolute path and, by default, does not restrict access to a workspace directory.
  • The MCP source allowlists `.hwp`, `.hwpx`, `.pdf`, `.xlsx`, and `.docx` files, resolves symlinks, and applies a 500MB full-parse file limit.
  • Metadata-only parsing has a smaller 50MB limit, while archive, PDF, HWP5, and table parsers have additional resource limits documented in the security policy.
  • Form filling can write Markdown or HWPX files to an `output_path`; require confirmation before allowing an agent to create or overwrite documents.
  • Kordoc processes untrusted binary formats; keep the Node.js runtime and parser dependencies updated and avoid running it on high-privilege accounts.
  • The CLI also has watch and webhook flows; keep those separate from the MCP server unless you have reviewed outbound notification behavior.

Privacy notes

  • Parsed content can include full document text, tables, titles, authors, page counts, metadata, warnings, form labels, and filled values.
  • Document paths and resolved output paths can appear in MCP responses, CLI output, logs, or saved artifacts.
  • Government, legal, HR, finance, or customer documents may contain sensitive identifiers that become Markdown or JSON in the MCP client context.
  • HWPX form filling can preserve and rewrite original templates; store generated files only in approved directories.
  • Webhook or watch-mode workflows can send converted document information to third-party endpoints if enabled outside the MCP server.

Prerequisites

  • Node.js 18 or newer.
  • A local MCP client that can launch stdio servers with access to the documents you want parsed.
  • Optional `pdfjs-dist` support if you need PDF parsing in environments where the package is not already installed.
  • Approved directories for source documents and generated Markdown or HWPX output.
  • Hancom Office on Windows if you rely on the repository's documented COM fallback for protected HWP or HWPX files.

Schema details

Install type
cli
Troubleshooting
No
Source repository stats
Scope
Source repo
Collection metadata
Estimated setup
10 minutes
Difficulty
intermediate
Tool listing metadata
Disclosure
Community-maintained MIT MCP server and document parser for Korean HWP, HWPX, HWPML, PDF, XLSX, and DOCX workflows.
Full copyable content
{
  "mcpServers": {
    "kordoc": {
      "command": "npx",
      "args": ["-y", "kordoc", "mcp"]
    }
  }
}

About this resource

Content

Kordoc MCP Server connects Claude and other MCP clients to the Kordoc document parser. It focuses on Korean document workflows where HWP, HWPX, HWPML, PDF, XLSX, and DOCX files need to become Markdown or structured metadata before an AI assistant can review them.

Use it when Claude needs supervised local access to parse documents, inspect metadata, extract a page range, pull one table, compare two versions, detect form fields, or fill an HWPX form template.

Source Review

These sources were reviewed on 2026-06-06. Prefer the live repository, English README, npm metadata, license, security policy, MCP server source, CLI source, and parser exports for current setup and behavior details.

Features

  • Parse HWP, HWPX, HWPML, PDF, XLSX, and DOCX files into Markdown.
  • Detect document format from file headers.
  • Extract document metadata without a full parse when supported.
  • Parse selected page or section ranges.
  • Extract a specific table by index.
  • Compare two documents across supported formats.
  • Extract structured form fields from Korean form templates.
  • Fill form fields and optionally write Markdown or HWPX output.
  • Return outlines, warnings, and quality signals when parser support is available.

Installation

Run the setup command from the upstream package:

npx -y kordoc setup

Then add the MCP server to your client configuration:

{
  "mcpServers": {
    "kordoc": {
      "command": "npx",
      "args": ["-y", "kordoc", "mcp"]
    }
  }
}

The npm package exposes both a kordoc CLI and a kordoc-mcp binary. The README recommends launching the MCP server through the kordoc mcp subcommand.

Use Cases

  • Convert Korean public-sector HWP or HWPX files into Markdown for Claude review.
  • Ask Claude to extract metadata, document outlines, warnings, or table content before summarizing a file.
  • Compare old and new versions of a document and identify added, removed, or changed blocks.
  • Pull fields from an application template and draft candidate fill values for user approval.
  • Fill an HWPX template and write the result to an approved output path.
  • Flag PDFs that likely need OCR before sending them through a separate OCR workflow.

Safety and Privacy

Kordoc MCP Server is a local document parser, not a sandbox. It resolves absolute paths and can read any allowlisted document file the server process can access, so run it under a least-privilege account and keep sensitive directories out of reach.

Treat parsed Markdown, metadata, form values, document paths, warnings, and generated HWPX files as sensitive. Review output paths before allowing writes, and avoid enabling watch or webhook workflows unless their network behavior is part of your data-handling plan.

#hwp#hwpx#document-parsing#korean-documents#office-documents

Source citations

Signals

Loading live community signals…

More like this, weekly

A short, calm digest of reviewed Claude resources. Unsubscribe any time.