Check ChatGPT's work, even when it edits fourteen datasheets at once

Changing a system's design pressure from 150 to 175 psig can mean updating every datasheet, design basis, and line list in a project. This is easy for ChatGPT or Claude - give it the background and it will churn through the updates in a few minutes.

The problem is checking every file that AI updated? Without Kord, you'd probably open every file and scan through for obvious errors and to check that it didn't make spurious changes. Maybe it reworded a materials call-out nobody asked it to touch, or just sounded obviously AI-written. Checking every file means you might as well have done the work yourself.

Kord shows every change that the AI made, with every document in one place. You can follow a single link and get the review done much, much faster.

Move beyond copy-paste

Most people use AI by copying into and out of a chatbox. Paste in a few paragraphs, get a rewrite, paste into your document. This is slow, but lets you sanity check what was happening.

Agents can now work on real files, and they move better and faster than copy-paste. Without a human in the loop, though, you don't know whether the changes are correct. And SharePoint's version history only tells you that a file changed, not what changed inside it.

Kord gives you the speed of AI agents with the safety of having a human in the loop.

How it works

Kord runs a Model Context Protocol (MCP) server. Claude Code, Claude Cowork, ChatGPT Work, Cursor, and more can connect with one URL and then work with your document sets. Your files stay where they are - Kord connects to your shared drive - and the agent edits documents in place. Kord then shows you exactly what the changes are.

Connecting with one URL and a browser sign-in, using your existing Google or Microsoft account. Your AI then talks to Kord, and can see the teams, repositories, and review sessions you have (and nothing else). ChatGPT and Claude let you control which actions a connector can take on its own, and you can revoke Kord's access at any time.

Either you or the AI can open a review session for the files that changed. The review session shows changes: line by line on text, cell by cell on spreadsheets, with changed text called out on PDFs and Word documents.

You can use Kord without a login - have the agent upload files to our MCP, and see the change immediately. Kord won't save the files - we just show the diff for a short time. This gives you some of the benefits without a login, install, or paying a cent.

• You see every change, not just what the agent reports. Edits in files you never asked about show up alongside the ones you expected.
• No opening two revisions side by side to hunt for three changed lines.
• Assign required reviewers, make comments and markups, and check that new revisions fix the problem.
• Every change is versioned, with an author and a timestamp.

Give agents the file history

The connection to Kord runs both directions. An agent can pull the text out of a file, walk its revision history, look at what is already under review and what the changes are. It can actually tell what Randy did last week.

• Ask Claude/ChatGPT to read the spec against the datasheets and flag where they disagree. It comments on the line or cell, inside the review session, where whoever fixes it is already looking.
• Check what changed last week, across a whole document set. Verify that everything was properly updated.
• Kord comments are threaded, resolvable, and written into a drive-linked Word file as a native Word comment.
• Let your AI tell you which review sessions are waiting on your approval, across every repository you have access to.

Sign-off stays with a person

An AI connected to Kord cannot approve or publish documents. Those are left out on purpose.

You can assign reviewers as required or optional - required reviewers block publishing until they approve; optional can weigh in without holding the release. A drawing that needs two disciplines and a QA check can have three required reviewers. A minor markup revision gets one. Either way, you can track exactly who approved what.

A client, auditor, or QA lead can pull up this revision in three years, then see an approval with a name, date, and the exact changes.

Let AI do the tedious work of making minor edits across fourteen datasheets, and let an engineer review a clear diff and put their name on it.

Setup is easy: a single URL and an in-browser sign-in. Kord works with ChatGPT Work, Claude Code, Claude Cowork, Cursor, and any other MCP client - the steps for each are at withkord.com/mcp.