VA DocParse Document Parser

Problem

Scanned PDFs and image-based documents often contain text, tables, and formulas that are hard to feed into downstream workflows. va-docparse targets this parsing scenario by calling a remote document parsing service through an MCP server. It converts PDF files, scanned PDFs, and common image formats into readable structured content, reducing the need for local OCR environments, custom parsing code, and multi-format recognition maintenance.

How It Works

The skill responds to intents such as OCR, document parsing, recognition, and reading, then calls the parse_document tool. The required input is an absolute file path, and the optional output_format can be markdown, json, or both. On success, the service returns extracted text content. On failure, it returns a coded error message, such as [F001] file not found, [F002] file unreadable, or [F003] unsupported format. For batch processing, each file is parsed independently; if the user requests merging, the results are concatenated in order with masked file names and failure codes preserved.

Boundaries

It supports PDF, scanned PDFs, PNG, JPG, and JPEG. It does not support audio, video, landscape photos, or person photos, and it does not directly support Word, Excel, or PPT; those files should be converted to PDF first. Results should be returned as the MCP service response without unsupported rewriting, completion, or expansion. When configuration is missing, network issues occur, or the service fails, the skill should report the error code and stop, without exposing MCP endpoints, keys, underlying commands, or other sensitive information.