# HWPX Automation MCP server

> Lets an AI client read, edit, create, and validate local HWPX documents through a workspace-scoped MCP server.

- Repository: https://github.com/airmang/python-hwpx-automation
- Page: https://hostedmcp.org/servers/airmang/python-hwpx-automation
- Transport: —
- Stars: 69
- License: Apache-2.0

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `HWPX_AUTOMATION_WORKSPACE_ROOTS` | no | no |  |

## FAQ

### How do I install the HWPX MCP server?

Install the package with the mcp extra and run the hwpx-automation-mcp command, or point your client at uvx --from "python-hwpx-automation[mcp]==7.3.1" hwpx-automation-mcp.

### Does it need Hancom Office or Windows to work?

No. The engine is pure Python, so it runs anywhere Python runs; Hancom is only needed if you want to render documents to PDF with the optional oracle extra.

### How do I restrict which folders the server can access?

Set HWPX_AUTOMATION_WORKSPACE_ROOTS to a JSON array of absolute or ~ paths; the server rejects traversal and symlink escapes outside those roots.

### Can an AI edit an existing HWPX document safely?

Yes, but the recommended flow is copy first, make the smallest edit, then re-read to verify, because edit tools save immediately.

### Is there a lightweight mode for small local models?

Yes. Use --profile minimal or set HWPX_AUTOMATION_PROFILE=minimal to expose only six core tools like read_outline, apply_text_changes, and validate_document.