GitHub announced on August 28, 2026 that Copilot in Visual Studio can review uncommitted changes or individual commits before a pull request is opened. The Git agent presents findings inline in the editor and in a navigable list inside Git Changes, where developers can continue the Copilot Chat conversation about each suggestion.

The feature is part of the August 2026 Copilot update for Visual Studio 2026 and works with both GitHub and Azure DevOps repositories. GitHub says the update is available across all Copilot plans, including Copilot Free.

What changed in Visual Studio

  • Review before the PR: ask the Git agent to inspect working-tree changes or a selected commit before opening a pull request.
  • Inline findings: review comments appear against the relevant code, with a findings list in Git Changes for navigation.
  • Conversational follow-up: continue in Copilot Chat to understand a finding or ask for help addressing it.
  • Repository support: the workflow supports repositories hosted on GitHub and Azure DevOps.
  • Broad plan access: GitHub lists the August update as available on all GitHub Copilot plans.

Why this matters for QA engineers

Copilot Git agent pre-PR review moves an AI feedback loop earlier in development. Test automation engineers can inspect a locator fix, assertion change, fixture refactor, or CI configuration edit before the branch creates reviewer noise. Earlier feedback can be useful, but it is still probabilistic review rather than a release gate.

  • Test the reviewer itself: seed known defects in a safe branch and measure detection, false positives, duplicate findings, and missed risks.
  • Compare review scopes: verify that working-tree review and commit review cover the files and changes your team expects, including renamed and generated test assets.
  • Require executable evidence: pair suggestions with unit, API, UI, static-analysis, or pipeline results rather than accepting a conversational explanation alone.
  • Protect sensitive context: check which diffs, logs, secrets, and local configuration can enter the Copilot interaction under your organization policy.
  • Keep independent gates: retain human review, branch protections, CI checks, and security scanning after a clean Git-agent result.

A practical QA pilot

  1. Create a non-production branch with one intentional functional defect, one weak assertion, and one harmless formatting change.
  2. Run the Git-agent review against the uncommitted working tree and record every finding.
  3. Commit the same change set, review that commit, and compare coverage and wording.
  4. Run the repository’s normal automated checks and compare their evidence with the AI findings.
  5. Have a human reviewer classify each finding as useful, duplicate, incorrect, or missed.
  6. Repeat on a representative automation change before defining team guidance.

Bottom line

GitHub’s August update gives Visual Studio users a convenient AI review checkpoint before a pull request exists. For QA teams, the value depends on whether the Git agent finds real defects without distracting reviewers or weakening established controls. Pilot it with seeded issues and measurable outcomes before treating it as part of the standard workflow.

Sources