How to Redline Word Documents in ChatGPT with Scaffold
Connect the Scaffold MCP to ChatGPT and redline Word documents with real tracked changes — step-by-step setup guide for professionals.
You can redline Word documents inside ChatGPT by connecting the Scaffold MCP connector to your ChatGPT workspace. Once connected, you upload your .docx file through Scaffold, ask ChatGPT to make specific edits, and download a revised document with real Word tracked changes — not a plain-text summary of changes, but an actual .docx with accept/reject markup.
This is not the same as uploading a file directly to ChatGPT. Native file uploads let ChatGPT read and summarize your document, but they cannot produce a .docx with tracked changes. The Scaffold MCP connector fills that gap.
What You Need Before You Start
- A Scaffold account — the Free plan covers everything in this guide
- ChatGPT Plus or Team (MCP connectors require a paid ChatGPT account)
- A Word document (.docx) you want to redline
You do not need to install anything on your computer. No Word add-in, no desktop agent, no browser extension.
How to Connect the Scaffold MCP Connector to ChatGPT
ChatGPT supports external tools through its connector system. The Scaffold MCP connector appears there once you authorize it.
Step 1 — Get your Scaffold connector URL. Create a Scaffold account at app.scaffoldyourdocs.com if you don't have one. The connector URL is the same for everyone: https://mcp.scaffoldyourdocs.com/mcp. There is no API key to copy.
Step 2 — Add the connector in ChatGPT. Open ChatGPT's settings and add Scaffold as a connector with the URL https://mcp.scaffoldyourdocs.com/mcp. When ChatGPT asks you to connect, verify your email with the code Scaffold sends and approve the connection.
Step 3 — Confirm the connection. Start a new ChatGPT conversation. In the tools panel, you should see Scaffold listed as an available connector with tools like upload_documents, scaffold_read, and scaffold_redline. If it does not appear, refresh the page and try toggling the connector off and on.
How to Redline a Document: Step by Step
Once the connector is active, the workflow is straightforward.
Step 1 — Upload your document. In ChatGPT, type something like: "Use Scaffold to upload my contract." ChatGPT will call the Scaffold upload_documents tool; attach your .docx when prompted. Scaffold keeps the document and its versions in your account until you delete them.
Step 2 — Give your redline instructions. Be specific. Rather than "improve this contract," say: "Add a mutual confidentiality clause after Section 3. Shorten the payment terms in Section 5 from 60 days to 30 days. Flag the indemnification clause in Section 8 as potentially one-sided and suggest alternative language." The more precise your instructions, the more useful the tracked changes will be.
Step 3 — Run the redline. Ask ChatGPT to apply the changes using Scaffold. ChatGPT calls the scaffold_redline tool, sends your instructions alongside the document, and Scaffold produces a revised .docx with each change recorded as a tracked change — shown in red strikethrough for deletions and underline for insertions, just like a manual redline.
Step 4 — Download and review. Use the Download button in the Scaffold pane to save the result, or ask ChatGPT to email it to you. Open the file in Word, turn on "Show Markup" in the Review tab, and review each tracked change. Accept or reject changes individually, or accept all at once if everything looks good.
How to Write Effective Redline Prompts for ChatGPT
The quality of the redline depends heavily on how you phrase your request. A few patterns that work well:
- Clause-level instructions: "In the Limitation of Liability clause, cap damages at two times the contract value rather than the full contract value."
- Structural changes: "Move the Governing Law clause from Section 12 to Section 2."
- Flagging concerns: "Mark any clause that waives consequential damages with a comment explaining the risk."
- Tone adjustments: "Soften the cure period language in Section 7 — make it less adversarial while preserving the substance."
Avoid vague requests like "make this better" or "fix the legal issues." ChatGPT performs best when instructions map to specific sections and specific changes.
Reviewing AI-Proposed Changes in Word
When you open the downloaded .docx in Word, use the Review > Track Changes panel. You can navigate change by change using the Previous/Next arrows and leave your own comments before sending the document to colleagues.
If a change looks wrong or overreaches, reject it. The underlying document stays intact. This is the same workflow you would use reviewing a human colleague's redline — the AI's suggestions are proposals, not instructions.
FAQ
What file formats does Scaffold support?
The Scaffold MCP connector works with .docx files (Microsoft Word Open XML format). It does not currently support .doc (legacy Word), .pdf, Google Docs, or plain text files. If your document is in another format, convert it to .docx in Word before uploading. PDF-to-Word conversion tools can handle most static PDFs, though complex layouts may need manual cleanup after conversion.
Does this work with the free version of ChatGPT?
No. MCP connectors in ChatGPT require a paid account, such as ChatGPT Plus or ChatGPT Team. The free tier does not support external tool connectors. If you are on the free tier and want to use Scaffold with an AI, Claude.ai's free plan supports MCP on desktop, which is an alternative path to the same result.
How is this different from ChatGPT's native file upload?
ChatGPT's native file upload lets the model read your document and respond with text — summaries, suggestions, or rewritten passages. What it cannot do is produce a .docx file with embedded tracked changes. That requires the document to be processed server-side by software that understands the Word XML format and can inject tracked-change markup. The Scaffold MCP connector does exactly that: it takes ChatGPT's instructions and applies them as real Word tracked changes, then returns a file you can open and review in Word.
Is my document stored on Scaffold's servers?
Yes. Scaffold keeps a document and its versions in your account until you delete them. In the Scaffold pane, More → Delete document removes the document and all its versions (this can't be undone); Delete version removes one version and keeps the others. Scaffold does not use your document content to train models. Review the current privacy policy at scaffoldyourdocs.com for specifics.
Can I compare versions alongside redlining?
Yes. Scaffold keeps every version of your document, with what went in and what was turned back. You can compare any two versions word by word, add a difference to the earlier version or revert it in the later one, and review ChatGPT's proposed changes one at a time on the Track, approving or rejecting each. Redlines and comparisons can be used in the same session.
Ready to try it? Start Scaffold free and open your first AI-redlined document in Word.