The Backstory

“I was told I could listen to the radio at a reasonable volume from nine to eleven…”

The Relocation

Milton Waddams was relocated to the basement. They took his stapler. But down there, surrounded by boxes of .doc files from 1997 and .xls spreadsheets that predate Unicode, he became something else entirely.

He became a document processing expert.

Fun Fact

After the movie Office Space came out, Swingline didn’t actually make a red stapler. Demand was so high they started manufacturing them. The movie created the product.

The Problem

Every enterprise has them:

• The Archive Folder — 50,000 Word documents from before the cloud existed
• The Legacy Database Export — Excel files with formulas referencing cells that no longer exist
• The Board Presentations — PowerPoint decks with embedded charts from 2003
• The Contract Repository — .doc files that crash modern Word

AI agents can read PDFs. They can parse JSON. But Office documents? The binary formats, the OLE containers, the OOXML with custom schemas?

Nobody wants to deal with that.

The Solution

mcwaddams handles the documents nobody else wants to touch.

# Extract text from a 1997 .doc file
result = await extract_text("contract_final_FINAL_v2.doc")

# It just works
print(result["text"])

What We Handle

Format	Era	Status
.docx	2007+	✅ Full support
.doc	1997-2007	✅ Works fine
.xlsx	2007+	✅ Full support
.xls	1997-2007	✅ Works fine
.pptx	2007+	✅ Full support
.ppt	1997-2007	✅ Works fine

Did You Get the Memo?

If a document is password-protected or encrypted, we’ll detect it and tell you. We can’t extract content from encrypted files, but we won’t silently fail either.

The Philosophy

1. No Silent Failures

When python-docx can’t handle a file, mammoth tries. When openpyxl fails, pandas steps in. You’ll always get either content or a clear error message explaining why.

2. Legacy is Not Abandoned

Those .doc files from 2003? They’re still business-critical for someone. We don’t treat legacy formats as second-class citizens.

3. Context-Aware Extraction

Large documents get paginated automatically. The MCP resource system lets you fetch chapters on-demand. Your context window stays manageable.

4. Testing Painful Stuff

We threw 301 random Office documents at mcwaddams. 299 succeeded. The 2 failures were empty/corrupt files.

See the TPS Reports for proof.

---

The Name

Milton Waddams. The guy with the stapler. Relegated to the basement with the old filing cabinets and the roaches.

That’s where the legacy documents live too.

“I could set the building on fire…”

---

Ready to start? → Installation