GitHub Expands Copilot Metrics Across Feature Use and CLI Customizations
New 28-day fields show which Copilot experiences and agentic CLI extensions are being used, with important counting and privacy limits.
Edited by Tyronne Panaino
GitHub expanded Copilot reporting on September 17 with two related changes for enterprise and organisation administrators. The impact dashboard and report APIs now break regular use down by Copilot feature, while the usage metrics API adds activity fields for agentic CLI customizations such as skills, custom agents, Model Context Protocol servers, slash commands and plugins.
The additions give administrators a more specific view than a single active-user total, but the definitions matter. These are product-usage measurements, not direct evidence that a feature improved code quality, delivery speed or business outcomes.
Feature engagement gets a 28-day view
GitHub defines feature engagement as an active user using an included feature on at least two days during a 28-day period. The dashboard and aggregate reports can break that activity down across code completion, agent edit, passive and active Copilot code review, the Copilot cloud agent, Copilot CLI and the Copilot app.
The new `copilot_feature_engagement` object carries the active-user total and counts by feature. A separate `users_in_phase_28d` field reports the full rolling 28-day population assigned to each AI-adoption phase as of the report day. GitHub retains `total_engaged_users` for people in that phase who were active on the individual day. That makes the two fields related but not interchangeable.
The feature-engagement release note also sets interpretation limits. One person can count under more than one feature. The feature-engagement object may be absent or null when GitHub cannot calculate it, while zero means the population was measured and had no matching users. The breakdown is available in enterprise and organisation 28-day aggregate reports rather than user-level reports.
CLI reports expose customization activity
The second update adds fields for skills, custom agents, MCP servers, slash commands and plugins across Copilot CLI reports. Arrays such as `totals_by_skill` and `totals_by_mcp` list as many as five items with the most recorded activity. Separate distinct-use counts include the wider variety of items used, not just those appearing in the top five.
The CLI metrics release note explains that customer-defined names are hidden: most appear under `other`, while custom slash commands appear under `custom`. MCP activity counts connection or reconnection attempts, successful or failed, rather than every tool call through an established connection. Plugin interactions also appear in skill totals, so adding those two totals would double-count part of the activity.
These boundaries make the fields useful for finding adoption patterns and enablement gaps, not for ranking tool effectiveness. Access requires an eligible enterprise or organisation role with the view-metrics permission, and the Copilot usage metrics policy must be enabled.
Status
Confirmed. GitHub documents both reporting changes as available. Confidence is medium because the definitions and behaviour come from the product owner and have not been independently validated against customer outcomes.
Sources
- GitHub Changelog — feature engagement in the Copilot impact dashboard
- GitHub Changelog — agentic CLI customizations in the usage metrics API
Update note: Last reviewed 2026-09-22. We will revise this post if GitHub changes the field definitions, report coverage or access requirements.
Sources
Drafted with AI assistance from source briefs; reviewed for citation completeness and label accuracy.