Copilot CLI usage metrics add agentic activity fields for enterprise reporting

GitHub Copilot has extended its CLI usage‑metrics API with a set of agentic activity fields that surface per‑user and aggregate data for skills, custom agents, Model Context Protocol (MCP) servers, slash commands and plugins. Enterprise and organization administrators can now query these metrics to see which custom extensions are being invoked and how diverse the usage is across their teams.

What’s New in the Metrics API

The API now returns the following arrays in 1‑day and 28‑day reports:

• totals_by_skill, totals_by_custom_agent, totals_by_mcp, totals_by_slash_cmd, totals_by_plugin – each lists up to five items with the highest interaction_count.
• distinct_skill_use_count, distinct_custom_agent_use_count, distinct_mcp_use_count, distinct_slash_cmd_use_count, distinct_plugin_use_count – each reports the number of unique items used.

In a per‑user report, an item contributes once to the distinct count; in an aggregate report the count reflects unique items used by any user, not a sum of per‑user totals. Empty arrays or zero values mean no activity, and the fields are omitted when the underlying data is unavailable.

Operational and Security Considerations

Only GitHub‑provided item names are displayed; customer‑defined names are masked to preserve privacy. Skills, custom agents, MCP servers and plugins appear under the other label, while slash commands are grouped under custom. For MCP servers the interaction_count increments on every connection or reconnection attempt, regardless of success, and repeated calls to the same server do not add additional counts.

Plugin metrics are a subset of skill metrics – every plugin‑driven skill invocation is counted in both totals_by_plugin and totals_by_skill, but skill‑only invocations appear solely in the skill totals. Adding the two totals together would double‑count plugin activity.

Access to these reports requires the View Copilot Metrics permission on an enterprise or organization role, and the Copilot usage‑metrics policy must be enabled. Administrators should verify that the appropriate roles are granted and that the policy is active before relying on the data.

How Practitioners Can Leverage the Data

With the new fields, teams can:

• Identify the most frequently used custom agents or skills and prioritize support or optimization for those.
• Track the growth of distinct item usage to gauge the breadth of automation adoption.
• Detect unexpected spikes in MCP connection attempts, which may indicate connectivity issues or mis‑configured agents.
• Correlate plugin usage with skill usage to understand the impact of plugin‑driven workflows.

Because the metrics are scoped to enterprise or organization boundaries, they can be incorporated into existing observability dashboards or used to drive capacity‑planning decisions for shared resources such as MCP servers.

Related CloudNinjas coverage: AI engineering.

What This Means For Practitioners

• Enable the Copilot usage‑metrics policy and ensure the View Copilot Metrics permission is assigned to the roles that need visibility.
• Ingest the new arrays into your monitoring or analytics pipeline; treat interaction_count as a raw event count, not a success metric.
• When evaluating plugin impact, remember that plugin totals are included in skill totals – avoid double‑counting in reports.
• Watch for empty or null fields as an indicator that a particular customization type is not being used or that telemetry is unavailable.