SquarePact - Run OpenClaw in Word
Native MS Word integration for your OpenClaw agent.
Start new thread

SquarePact - Run OpenClaw in Word - Native MS Word integration for your OpenClaw agent.

by•
If you’re a Word power user, you’ve likely built agentic workflows with OpenClaw. But when reviewing or drafting Word documents, the workflow falls apart. You manually copy and paste text into a chat interface, lose your formatting, and spend time fixing numbered lists, headers, and styles. SquarePact brings your OpenClaw agent directly into Word. Because it understands Word's native structure, your agent can read the document, make changes, and reconcile text without destroying your formatting.

Add a comment

Replies

Best
Maker
📌
Hey Product Hunt! 👋 I’m excited to introduce SquarePact. We built SquarePact to solve the "last mile" problem in AI document workflows. If you are a power user, you’ve probably already built an agentic workflow using OpenClaw. But when it actually comes time to review, draft, or edit a complex Word document, the process hits a wall. You end up copy-pasting text out of a chat UI into Word, and suddenly your numbered lists reset, your styles are wiped out, and you spend 20 minutes playing formatting janitor. Most AI tools try to force you into their walled garden. We took the opposite approach: SquarePact is a native front-end UI for your custom AI workflows. Here is what you can do with it: - Direct OpenClaw Integration: Bring your OpenClaw agent directly into the Microsoft Word ribbon. - Zero Formatting Drift: SquarePact edits word documents natively. Your agent can read the document, make changes, and reconcile text without destroying your styles, headers, or tables. - Total Privacy: Route entirely to local models or your private cloud. Your document content stays within your own infrastructure. - Beyond Spellcheck: Use our Scrivener's Error sequence to catch transposed numbers, reconcile definition mismatches, and fix broken cross-references. We built this for the developers, tinkerers, and professionals who want to use their own models but need them to actually function inside the tool where the real work happens. We'd love to hear your feedback, answer any questions, and hear how you're currently handling the hand-off between your local AI agents and your Word documents!

This is exactly what I needed for editing lengthy contracts. OpenClaw preserved my numbered lists and headers perfectly, no fixing formatting afterward.

I've been following this project a while on LinkedIn from . Even with tools like Gemini for Google Doc, Copilot for Word and ChatGPT, Claude, etc. in general, I'm still having to look through the edit history to see what changed and what didn't. This is a very powerful tool to have as a side kick, very curious to start trying the more advanced agentic features! Great work!

Maker

 thanks Jacob, please feel free to reach out to me any time with your thoughts (positive and negative)!

The formatting-preservation angle is such an underrated detail, most agent tools just bulldoze styles and leave you redoing numbered lists by hand. Glad someone finally built this with Word's native structure in mind instead of treating the doc like plain text.

How does the agent handle complex tracked changes or comments already in the document when making edits, and is there any risk of it clobbering those existing review states?

Maker

 you can see both! SquarePact can send your agent the raw xml which shows both the deleted and inserted components of tracked changes. Then your agent can handle it as you see best. Typically we find it best to just use the newest version.

The fact that it preserves native Word structure instead of just treating the doc as plain text is exactly the right call. So many tools get that wrong and end up mangling numbered lists and styles every time.

How does it actually handle complex tables and tracked changes when the agent edits the doc, does it preserve all the formatting nuances or just the basics?

Maker

 SquarePact can send the raw xml if your agent requests it, so all table and tracked changes info is available. Tables are notoriously tricky though, and we find that agents planning on doing heavy table work should use at least the equivalent of Opus or higher.

How does it handle really long documents where the agent might need to make dozens of small edits across different sections, does it work on the whole file at once or only selected ranges?

Maker

 great question! That's why SquarePact allows your agent to work directly inside of Word: it can surgically edit individual paragraphs (and even individual words) rather than needing to rewrite the entire document at once each time. That way the edits show as track changes as well!

How does it actually handle tracked changes or comments from reviewers when the agent rewrites a paragraph, does it leave those intact or just blow through them?