GitHub adds repository-level Copilot usage metrics for enterprise and organization admins

GitHub has made repository-level Copilot usage reporting generally available, giving enterprise and organization administrators a more granular way to see where AI-assisted development is affecting pull request work. The July 17 changelog says the Copilot usage metrics REST API now includes two new one-day report endpoints that break activity down by repository rather than stopping at the enterprise, organization or user level.

The new reports cover pull requests created and merged by Copilot coding agent, along with pull requests reviewed by Copilot code review. For code review activity, GitHub says the response includes suggestion counts broken down by comment type. The company positions the change as a foundation for repository insights and AI-readiness reporting, because teams can identify which codebases are already seeing activity from Copilot automation and where additional enablement might have the biggest effect.

Why it matters

AI coding tools are moving from individual productivity experiments into managed engineering programs, and that shift puts more pressure on measurement. Organization-wide totals can show adoption, but they do not tell platform teams whether Copilot is being used in a core service, a low-risk internal tool, a legacy repository or a team that is still evaluating the workflow. Repository-level reporting gives administrators a clearer basis for coaching teams, monitoring rollout patterns and comparing use against the maturity of each codebase.

The endpoints are also narrow enough to matter operationally. GitHub lists separate paths for enterprise and organization reports: one under /enterprises/{enterprise}/copilot/metrics/reports/repos-1-day and another under /orgs/{org}/copilot/metrics/reports/repos-1-day, each accepting a specific day parameter. That makes the data easier to fold into existing dashboards, governance reviews or internal developer platform reporting without waiting for a monthly aggregate.

GitHub notes that the reports focus on Copilot coding agent and Copilot code review pull request activity. That means the data should be read as a view into repository-level automation and review workflows, not as a complete measure of every Copilot interaction across chat, IDE completions or other surfaces. Even with that boundary, the release is a useful signal: vendors are responding to enterprise buyers that want AI coding tools to come with auditable usage data, not just seat counts and productivity claims.

Sources

• github.blog

Cover photo by Daniil Komov on Pexels, used under the Pexels License.