About multi
A collaborative writing tool where people and AI agents work in the same document, built for the way teams really write with AI.
Why we built multi
Most AI writing tools are made for one person. The moment a team wants help from AI, someone has to leave the shared document, work with an assistant somewhere else, then bring the result back and explain it to everyone. The AI's contribution becomes invisible, and coordinating it becomes a chore.
multi takes a different approach: it puts AI agents directly into the collaborative writing environment. Agents edit the same document as the rest of the team, and their work stays visible to everyone through two shared objects, agent profiles and tasks, while their responses appear right in the familiar comment feature.
The research behind multi
multi began as a research prototype, built to study how teams write together with AI agents. We used a prototype of multi to conduct a study for the paper, “Collaborative Document Editing with Multiple Users and AI Agents”, which followed 14 teams over a week of real writing work. Interaction logs and interviews showed that teams don't treat agents similar to human team members but integrate them into their existing norms of authorship, control, and coordination. The agents people created and the outputs they produced became shared resources. In particular, users created multi agents to gain different perspectives on their writing.
We learned from this study what is important for collaborative document editing with AI: keeping AI use transparent, keeping control with the people, and that agent generated text becomes a shared resources for the team.
Read the full paper on ACM Digital Library or as a preprint on arXiv.
The team
Multi is the result of a research project that started in 2021. The researchers behind the tool are: Florian Lehmann, Krystsina Shauchenka, and Daniel Buschek, working at the intersection of human-computer interaction and AI-assisted writing.
Try it
The app is free to use. Start writing with your team and their agents in one shared workspace.