GitHub's Copilot code review switches Lite reviews to an ensemble of agents, reporting 47% more high-severity comments addressed at 8% lower cost

GitHub Changelog·high signalXBlueskyLinkedInCopy link

A 2026-09-11 GitHub Changelog entry says multiple agents now collaborate on Lite-level Copilot code reviews, with testing showing a 47% increase in addressed high-severity comments, 31% for medium and 11% for low, while cutting review cost roughly 8%. The review agent also gained the full Copilot SDK shell toolset, so it can run builds, execute tests and run scripts during review. Copilot now auto-resolves its own comments when a pushed commit addresses them, and autofix commits get contextual messages instead of a generic default.