Three agents — Detector, Writer, Reviewer — fix documentation; confident fixes become merge requests, uncertain ones become issues for humans.
The interesting decision here is knowing when to stop: confident fixes become merge requests, uncertain ones go back to a human as an issue. When agent demos are usually scored on how much they automate, drawing an explicit line around what not to do reads as maturity. It also lands its output in GitLab's native units, so the work slips into an existing review culture instead of asking teams to adopt a new surface.
Recent winners are built as teams of specialized agents, not single chatbots.
No new interface. Merge requests, issues, Slack, Teams, email — winners return their results into containers that already exist.
Projects that explicitly refuse to let a generative model decide tend to score higher, not lower.
Summaries are written by this site. Project pages include demo videos, screenshots and the team's own write-up (videos may autoplay).