GitHub expands the Copilot Usage Metrics API with code review metrics

GitHub has expanded the Usage Metrics API for Copilot with a new object, copilot_code_review, which appears in daily repository-level reports repos-1-day for enterprises and organizations. The report now includes the number of comments from the Copilot code review tool that were evaluated, and how many of them developers actually resolved, whether by accepting the suggestion, asking Copilot to fix it, manually editing the code, or replying that the fix will come in a subsequent pull request. Dividing these two numbers yields the so-called addressed rate for a given repository.

The metrics are further broken down by comment severity, review level, comment type, and customization used, such as custom instructions, skills, or MCP servers. The report also includes the field total_merged_no_human_change_request_after_review, which counts merged pull requests where, after review by Copilot, no human requested any further changes.

According to GitHub, the existing fields in the reports only showed how often Copilot code review ran and how often someone accepted a suggestion with a click, but did not capture the other ways developers respond to the feedback. The new reports are available via the API to users with View Copilot Metrics permission and require the Copilot usage metrics policy to be enabled. If a repository had no Copilot code review activity on a given day, the counts are zero and the fields are empty.