M
MCP Directory CURATED DIRECTORY
911 MCP servers indexed
中文
← Back to directory
K

Kordoc

Community
Convert Korean documents (HWP, HWPX, PDF, Office) to Markdown with CLI and MCP server, featuring form filling and diff.
How to install ↓ Compare GitHub source repository ↗
Category
Other
Stars
★ 2.4k Very popular
Transport
stdio (local process)
Runtime
Node.js 18+
Credentials
No credential needed
License
MIT
Last commit
Tools
15
59FMRS · C

Kordoc is a powerful document processing tool especially adept at handling Korean government documents. It offers broad format support, a rich set of MCP tools, and high customizability. For AI workflows requiring automated processing of Korean documents, it is a reliable choice, but its name and documentation language may not suit non-Korean users.

Strongest · Documentation 14/20 Weakest · Reliability 9/20

Reliability
9/20
Security and permissions
11/20
Maintenance
13/20
Documentation
14/20
Setup experience
12/20
Why each score
Reliability 9/20
This is a feature-rich MCP server claiming to support parsing HWP, HWPX, PDF, etc., and converting to Markdown. However, static review cannot verify the MCP handshake or whether declared tools match actual behavior, as no explicit MCP server tests or CI run evidence is visible in the repository. The README details 15 tools and common issues, but lacks reproducible test cases or CI configuration to prove key paths. Reliability score is capped at 12; given the lack of executable verification evidence, 9 is awarded.
Security and permissions 11/20
No red-flag signs of malware, credential theft, or covert exfiltration. The README mentions several security protections such as ZIP bomb protection, path traversal blocking, and SSRF prevention, but their implementation cannot be statically confirmed. Potential attack surface exists due to parsing PDF, ZIP, etc. Additionally, PDF OCR downloads external models, but KORDOC_OFFLINE and KORDOC_ROOT environment variables are supported to limit network and file access, showing some security consideration. Since permission and confirmation mechanisms are unclear and data boundaries mainly rely on optional environment variables, a score of 11 is given.
Maintenance 13/20
Repository is active with recent v4.7.2 releases and not archived. High star count (1672) indicates community interest, but per rules stars do not count. MIT license present, CHANGELOG is detailed, but no clear dependency update policy or security response channel. Maintenance is active but governance has gaps, such as lack of explicit contributing guide or security advisory, hence 13.
Documentation 14/20
Documentation is very thorough, covering installation, CLI, API, MCP config, usage examples, and changelog. But missing a dedicated troubleshooting guide for the MCP server itself, and full parameter descriptions for tools. Some hidden assumptions exist, like dependency on Node.js 18+, and OS-specific compatibility details not fully specified. Documentation is high quality but has minor omissions, so 14.
Setup experience 12/20
An interactive install wizard via `npx kordoc setup` supports various AI clients, and manual configuration examples are provided. Installation steps are clear, but depend on npx and Node.js, without mentioning version compatibility issues. Since actual installation cannot be verified and no CI evidence, setup is capped at 15 per guidelines, giving 12.

Static review · not runListed 2026-08-07

Read the FMRS scoring method →

Fit and risk

What it can accessReads local filesWrites / deletes local filesUses the network

Best for

  • Developers or teams needing to process Korean official documents like HWP and HWPX.
  • Users who need to parse and generate official documents in AI-assisted workflows.
  • Automation tasks involving a large number of Korean government or public sector documents.

Not for

  • Users needing non-Korean document formats (e.g., pure English Office docs) – although supported, it's not the primary focus.
  • General MCP use cases not requiring document conversion or heavy document handling.

Required permissions

  • Read and write local files (document parsing, generation, patching).
  • Network access: downloads OCR model (~18MB) on first use.
  • May read images (PNG/JPG/WebP) for OCR.

Risks and side effects

  • Processing untrusted files may present security risks, but the project includes security hardening like ZIP bomb and XXE protections.
  • OCR model download requires network access and may fail in offline environments.
  • Incorrect document parsing could lead to data loss; it's recommended to validate important documents.

Setup

Before you start

Runtime：Node.js 18+

Other optional settings (2)
KORDOC_OFFLINE optional Set to 1 to block all outbound requests (OCR model downloads, watch webhooks) before they are sent, for closed-network deployment; opt-in, default behavior unchanged when unset.
KORDOC_ROOT optional Set to a directory to restrict all MCP server file reads/writes to under that directory; opt-in, unrestricted when unset.
  1. Ensure Node.js 18+ is installed.
  2. Run npx -y kordoc setup and select your AI client (automatic configuration).
  3. Or manually edit your MCP client configuration file to add:
{"mcpServers":{"kordoc":{"command":"npx","args":["-y","kordoc","mcp"]}}}

For Windows, you may need cmd /c npx wrapper.

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

Shown for Claude Desktop. Other clients may use a different file or key (VS Code uses "servers") — the configurator below converts it.

.vscode/mcp.json
{
  "servers": {
    "kordoc": {
      "command": "npx",
      "args": [
        "-y",
        "kordoc",
        "mcp"
      ]
    }
  }
}

Goes in your project's .vscode/mcp.json (VS Code uses a "servers" key).

Terminal
claude mcp add kordoc -- npx -y kordoc mcp

Run it in a terminal; replace any <…> placeholders with your own values first.

Generate and validate client config

Check that it works

After running `npx -y kordoc setup` (or adding the mcpServers config manually) and restarting the client, check that tools like parse_document, fill_form, and generate_document appear in the tool list; or simply ask 'Parse this .hwpx file into Markdown' — a Markdown reply confirms the connection.

Troubleshooting

  1. If `npx -y kordoc setup` fails on Windows, use Command Prompt (cmd) instead of PowerShell, or set execution policy to RemoteSigned.
  2. If you encounter `MODULE_NOT_FOUND`, first run `npm uninstall -g kordoc`, then re-run the setup.
  3. If PDF parsing yields garbled text, try enabling OCR: add `ocr: true` parameter in MCP calls.
  4. For corrupted HWP files, Kordoc may not be able to fix them; ensure original files are not damaged.

Things to try

Once connected, you can ask your AI assistant things like:

  • Parse this HWPX file into Markdown and list the tables it contains.
  • Compare these two documents and generate an old-vs-new diff table.
  • Fill in the blank fields of this application form using the built-in gian template.
  • Apply my edited Markdown changes back to the original HWPX file while preserving formatting.

Tools 15

parse_document read-only
Parse HWP/HWPX/PDF/XLSX/DOCX documents to Markdown with metadata.
detect_format read-only
Detect file format using magic bytes.
parse_metadata read-only
Extract only the metadata of a document.
parse_pages read-only
Parse a specific page range of a document.
parse_table read-only
Extract the Nth table from a document.
compare_documents read-only
Compare two documents, supporting cross-format comparison and generating diff reports.
parse_form read-only
Extract form fields from a document as JSON.
fill_form writes
Fill values into a form template while preserving original HWPX formatting and integrity guards.
Show 7 more tools
patch_document writes
Apply edited Markdown patches to the original HWPX or HWP file, preserving formatting.
extract_profile read-only
Extract a table format profile from a reference HWPX, reusable for document generation.
generate_document writes
Generate an HWPX file from Markdown, including tables, formulas, and charts, with official document presets.
place_seal writes
Place a seal or signature image over anchor text for automatic stamping.
render_document read-only
Render HWPX to PNG or SVG preserving original layout, supporting multi-page and search highlights.
redact_document writes
Detect and mask PII such as resident numbers, phones, emails, cards, and bank accounts while preserving formatting.
parse_chunks read-only
Generate structured text chunks for RAG, preserving headings and hierarchy.

Use cases

Convert Korean government office documents like .hwp and .hwpx to Markdown for AI processing.
Automatically fill official forms and documents such as reports and plans.
Compare document versions to generate diff reports with added/removed/modified content.
Convert AI-generated Markdown back into standard-format HWPX official documents.
Redact documents containing personal information.

Supported clients

Claude Desktop
Cursor
Windsurf
Claude Code
VS Code
Gemini CLI
Zed
Codex

Listed from the project's documentation, not tested by this site.

Overview

Kordoc is a CLI and MCP server specialized in handling Korean official documents. It can parse HWP 3.x/5.x, HWPX, HWPML, PDF, XLS, XLSX, DOCX, and images (PNG/JPG/WebP) into Markdown. It offers form filling, document diff, format-preserving patches, Markdown to HWPX generation, PII redaction, and more. Developed by a former public official from Korea, it has been extensively tested against real government documents.

Similar servers

Telegram MCP Server 70 · B

Telegram MCP server powered by Telethon to let MCP clients read chats, manage groups, and send/modify messages, media, contacts, and settings.

★ 1.8k · Tools 76 Compare with this →
Compare with other MCP servers

Source revision 9780e501385b Data synced 2026-10-12 Read the FMRS scoring method