The latest GitHub update adds full API control over Copilot code review configuration and delivers a tighter review overview UI. Engineers can now script policy changes, integrate settings with CI pipelines, and obtain richer attribution data, while the more compact comment format reduces noise in pull‑request discussions.
Copilot code review API: configuration at scale
Both the REST and GraphQL interfaces now accept the same effort levels shown in the web UI and expose fields that identify the originating model, skills, MCP servers, and any custom instructions used for a comment. Role‑based access is enforced:
- Enterprise admins can toggle code review globally for all organizations; these enterprise‑managed settings cannot be overridden downstream.
- Organization admins control enablement, default effort, and runner configuration for their orgs.
- Repository admins adjust review effort and decide whether custom instructions or MCP tools are applied.
Enterprise and organization endpoints reject GitHub Apps and fine‑grained personal access tokens, while repository endpoints accept both token types provided the caller has Administration: read or Administration: write scopes. Classic PATs require repository‑admin rights.
Compact review overview reduces PR clutter
The new overview comment consolidates effort metrics and any survey prompts into a footer, merges skipped and excluded files into a single “Files not reviewed” block, and omits empty sections. Long file lists are collapsed with an overflow count. Reviewers can now provide negative feedback directly on the overview, selecting a reason such as a disagreeable approval assessment, overly verbose comment, or incorrect finding. This feedback loop is intended to improve future suggestions.
Operational and reliability enhancements
Several reliability fixes accompany the API rollout:
- Billing and preview settings persist when code review is toggled off and on.
- Failed or timed‑out reviews are automatically retried once, and the system can fall back to an alternate model if the preferred one is unavailable.
- GitHub Actions, Dependabot, and internal apps regain the ability to request reviews on bot‑authored pull requests when the “block reviews with external license” flag is enabled, provided the organization pays for the review.
- False failure reports at review start are eliminated, and in‑progress reviews receive extended time during service updates.
- Survey prompts are suppressed for organizations that have opted out of feedback collection.
Additional UI refinements include clearer wording on settings pages, automatic collapsing of findings unrelated to recent changes, and balanced‑effort reviews that now employ three distinct reviewer models plus a combined model.
Related CloudNinjas coverage: AI engineering.
What This Means For Practitioners
Teams should audit existing Copilot code review policies and migrate any manual configuration into the new API endpoints to enable version‑controlled, repeatable changes. When integrating with CI/CD, ensure the service account used has the appropriate Administration scopes for the target level (enterprise, org, or repo). Monitor the attribution fields to verify that the expected model and custom instructions are being applied, especially in regulated environments where provenance matters. Finally, test the fallback behavior by simulating model unavailability to confirm that reviews continue without manual intervention.
