doc-mcp

doc-mcp

An MCP server for extracting text content from documents.

Supported Formats

• PDF - Portable Document Format
• DOCX - Microsoft Word (Open XML)
• DOC - Microsoft Word (Legacy)
• PPTX - Microsoft PowerPoint (Open XML)
• ODT - OpenDocument Text
• TXT - Plain text

Installation

go install github.com/taigrr/doc-mcp@latest

Usage

With Crush

Add to your Crush config:

{
  "mcp": {
    "doc-mcp": {
      "command": "doc-mcp"
    }
  }
}

Tools

All tools support pagination to manage large documents and avoid filling context.

crack_file

Extract text content from a local file.

Input:

• path (string, required): Path to the document file
• max_chars (int, optional): Maximum characters to return (default 50000, use -1 for unlimited)
• page (int, optional): Page number for paginated results (1-indexed, default 1)
• page_size (int, optional): Characters per page (default 10000)

crack_url

Download and extract text content from a URL.

Input:

• url (string, required): URL of the document to download (http/https only)
• max_chars (int, optional): Maximum characters to return (default 50000, use -1 for unlimited)
• page (int, optional): Page number for paginated results (1-indexed, default 1)
• page_size (int, optional): Characters per page (default 10000)
• max_download_mb (int, optional): Maximum download size in MB (default 10, max 100)

crack_base64

Extract text from base64-encoded document data.

Input:

• data (string, required): Base64-encoded document content (standard or URL-safe)
• max_chars (int, optional): Maximum characters to return (default 50000, use -1 for unlimited)
• page (int, optional): Page number for paginated results (1-indexed, default 1)
• page_size (int, optional): Characters per page (default 10000)

Defaults

License

0BSD