GitHub's July 17 changelog is a small product update with a larger operating lesson. Repository-level Copilot usage metrics are now generally available through the Copilot usage metrics REST API. Two new endpoints return a daily, per-repository breakdown of pull request activity for Copilot coding agent and Copilot code review at both enterprise and organization levels.
The reported activity includes pull requests created by Copilot coding agent, pull requests merged by Copilot coding agent, pull requests reviewed by Copilot code review, and Copilot code review suggestion counts by comment type. GitHub says this lets teams see where Copilot coding agent and code review are driving pull request activity across a codebase, forming a foundation for repository insights and AI-readiness reporting. Access is tied to enterprise or organization ownership, billing manager roles, or custom roles with View Copilot Metrics, and the Copilot usage metrics policy must be enabled.
Grey Haven's read: AI coding agents are entering the management layer. The early adoption question was whether developers liked the assistant. The production question is which repositories are absorbing agent work, where review suggestions are concentrated, and whether generated pull requests are being merged safely. Repository-level reporting matters because software risk is not evenly distributed. A toy repo, a core payments service, and an old internal admin tool should not be governed with the same assumptions.
Operator guidance: engineering leaders should attach agent metrics to repository criticality. Track not only volume, but where agent-created pull requests land, how often humans change them, which repos receive repeated security or quality suggestions, and whether cycle time improves without increasing rollback or incident rates. Seat adoption is a vanity metric. Repo-level contribution, review burden, and defect outcomes are closer to the truth.
