--- title: mcwaddams description: MCP server for Microsoft Office document processing. Named for Milton Waddams, who was relocated to the basement with boxes of legacy documents. template: splash hero: tagline: "I was told there would be document extraction." image: file: ../../assets/stapler.svg actions: - text: Get Started link: /installation/ icon: right-arrow variant: primary - text: View on GitHub link: https://github.com/ryanmalloy/mcwaddams icon: external variant: minimal --- import { Card, CardGrid, LinkCard } from '@astrojs/starlight/components'; ## The Situation You have Office documents. Lots of them. Some are modern `.docx` files. Others are `.doc` files from 1997 that predate Unicode. Your AI agent needs to read them, but native PDF-to-text tools just stare blankly. **mcwaddams** handles what nobody else wants to touch. Extract text, tables, images, metadata. Analyze structure. Convert to Markdown. Index for on-demand fetching. `.doc`, `.xls`, `.ppt` — the formats everyone forgot about until they need them. We didn't forget. When python-docx chokes, mammoth steps in. When openpyxl fails, pandas takes over. No silent failures. Index once, fetch chapters/sheets/slides on demand. Keep your context window manageable. --- ## Quick Install ```bash # With uvx (recommended) uvx mcwaddams # Add to Claude Code claude mcp add mcwaddams "uvx mcwaddams" ``` --- ## Documentation Structure This documentation follows the [Diátaxis framework](https://diataxis.fr/): --- ## TPS Reports > *"Did you get the memo about the TPS reports?"* We take testing seriously. Check out our [Test Dashboard](/tps/dashboard/) for live results, or browse the [Torture Tests](/tps/torture/) where we threw 301 random Office documents at mcwaddams. **Result:** 99.3% success rate. The 2 failures were empty/corrupt files. ---
Named for Milton Waddams, who was relocated to the basement with the legacy documents.
"I could set the building on fire..."