Open the tool · See an example
When a group of people uses AI to draft a document together, the final version comes out clean and finished. But the record of how it got there disappears. Who asked for what? Whose feedback changed which section? Where did two reviewers disagree, and how was it resolved? That context lives in email threads, chat messages, and people's heads. Six months later, when someone asks why paragraph four says X instead of Y, nobody remembers.
Trace Collaborate keeps that record. It tracks each reviewer's feedback across every round of revision and maps it to the specific sections it changed. The output is a single document you can open in any browser that shows the full provenance of the final text: who influenced what, when, and why.
It extends Trace, which does the same thing for a single author working with AI. Trace Collaborate brings multiple reviewers into that loop.
Reviewers never touch this tool. They just review a Word document and leave comments, the same way they already do. The tool reads those comments and does the rest.
Here's the full cycle:
- Write your first draft. Open any AI tool you like and write the initial prompt. Paste the AI's output into Trace Collaborate along with a project title and the prompt you used.
-
Send the draft out for review. Save it as a .docx and share it however you normally would. Email, a shared drive, a chat message.
-
Reviewers add their comments in Word. They open the document, leave comments on specific sections or on the document as a whole, and send it back. Nothing changes about their workflow.
-
Drop the reviewed document into the tool. Open Trace Collaborate in your browser, drag the .docx file onto the page. The tool reads every comment and shows you who said what, anchored to the sections they were commenting on. You can also add feedback manually if someone gave it to you over email or chat instead of in Word.
- Copy the re-prompt. The tool packages all the feedback into a single block of text, organized by reviewer, with instructions for the AI to produce a revised draft and report what it changed and why. Copy that text and paste it into your AI tool.
-
Paste the AI's response back. The AI gives you a revised draft along with an attribution record. Paste the whole thing back into Trace Collaborate. The tool parses it, diffs it against the previous version, and records which reviewer's feedback drove each change.
-
Repeat. Send the new draft out for another round of review. Drop the new comments in. Generate another re-prompt. Keep going until the group converges.
-
Export. When you're done, export a Trace Collaborate document. It's a single HTML file you can open in any browser or share with anyone.
Nothing new. They receive a Word document, open it, leave comments, and send it back. They never need to visit a website, create an account, or install anything.
The exported document has three views:
Unit view shows the final text paragraph by paragraph. Click any paragraph to see its full history: who created it, which reviewer's feedback changed it in which round, and what they said. Each reviewer is color-coded so you can see influence at a glance.
Session view shows the complete chronological arc of the drafting process. Every round of reviewer feedback, every AI iteration, in order. You can see how the document evolved from the initial prompt to the final version.
Reviewer view shows a per-person summary. For each reviewer: how many comments they made, how many rounds they participated in, which sections they influenced, and what they said.
The document answers questions like: whose feedback shaped section three? Did everyone actually weigh in, or did one person's comments dominate? Where did two reviewers give conflicting input, and what happened? Which sections stabilized early and which ones kept changing?
The tool uses three layers to trace feedback through revisions:
The re-prompt asks the AI to report what it changed and who drove each change. The tool also diffs each revision against the previous version to independently verify what actually moved. And it knows which Word comments targeted which sections, so if the AI's self-reporting is incomplete, the tool can infer attribution from the comment anchors.
These three signals are cross-referenced at each round and accumulated into a provenance chain. Every paragraph in the final document carries its full history through every iteration.
No accounts are required. Project state is saved in your browser's local storage so you can close the tab and come back later.
Open the example to see a finished Trace Collaborate document. The example shows a sustainability FAQ drafted with AI and reviewed by three people across three rounds. Or go straight to the tool and start a project.
index.html The tool
viewer.html Opens exported Trace Collaborate documents
examples/ Example gallery and worked examples
lib/ Bundled dependencies (JSZip for reading .docx files)
No build step. No install. Open index.html in a browser and it works.
H/T Lucky Vidmar for the idea to extend Trace!
MIT