GitHub Copilot Code Review: Batch Apply Feedback with Cloud Agent

Closing the Loop Between Code Review and Implementation

GitHub has improved the experience of acting on Copilot code review feedback, making it significantly faster and more flexible to turn review suggestions into committed code. The update, released on May 19, 2026, centers on two changes: an enhanced "Fix with Copilot" dialog and a new batch processing capability.

Enhanced "Fix with Copilot" Dialog

The previous "Implement suggestion" button — which passed a single Copilot code review comment to the cloud agent — has been redesigned. Clicking the updated "Fix with Copilot" button now opens a dialog that gives developers meaningful control over how the fix is applied:

Destination branch choice. Developers can decide whether the cloud agent pushes changes directly to the current pull request branch or creates a new pull request against it. This is useful when the fix is simple and self-contained (apply directly) versus when it touches broader scope and warrants separate review.

Model selection. The dialog exposes the model picker, allowing developers to select which AI model the cloud agent uses for the fix. A lightweight model like Claude Haiku 4.5 or GPT-5.4-mini may be appropriate for a simple rename, while a more capable model might be chosen for a logic refactor.

Optional guidance. Developers can add free-text instructions to steer how Copilot implements the suggestion — for example, asking it to follow a specific naming convention or avoid touching unrelated lines.

Batch Processing Multiple Review Comments

The larger productivity unlock in this update is the "Fix batch with Copilot" button, which lets developers select multiple Copilot code review comments and address them all in a single cloud agent session. Previously, each suggestion had to be routed to the agent individually, requiring separate confirmation steps for every comment. On pull requests with a dozen or more review suggestions, this was a meaningful source of friction.

With batch processing, developers can queue up several suggestions at once — for example, all variable naming issues, all missing null checks, or all style violations — and hand them off to the cloud agent together. The agent handles the fixes in one pass, then pushes the result for review.

Context: A Faster Code Review Feedback Loop

This update builds on a series of improvements to Copilot code review made throughout early 2026. The review system was rebuilt on an agentic architecture in March 2026, and later gained severity labels and comment grouping (May 12, 2026). The May 19 update completes the feedback loop by making it as frictionless as possible to act on the suggestions the review system produces — not just read them.