DocMason
A repo-native agent that turns private office documents such as decks, spreadsheets, PDFs and emails into a local, evidence-first knowledge base, preserving structure like slide layout, table references and formatting so an AI agent's answers stay traceable to a source file.
DocMason compiles private work files such as decks, spreadsheets, PDFs and emails into a local, file-based knowledge base with provenance, so that a coding or research agent reasons over structured, multimodal evidence rather than a single unstructured text blob. The stated goal is that every answer can be traced back to where it came from in the source documents.
It targets problems other document-AI tools tend to lose: slide layout, presenter notes and chart-to-text relationships in decks, multi-sheet references and nested tables in spreadsheets, formatting used as meaning (such as red text marking risk), and reasoning that spans multiple related documents. It enforces data contracts and provenance boundaries, and validates the resulting knowledge base against code rules, all locally.
It is operated through ChatGPT Work, Codex mode or Claude Code rather than through a separate cloud service, and nothing leaves the local machine during processing.