This article is a technical explanation and implementation example generated using AI. The code and procedures presented are based on primary sources, but have not been verified by the author on actual hardware. Behavior may vary depending on the environment and version.
A new reporting scope regarding Agentic CLI customizations has been added to the Usage Metrics API for tracking GitHub Copilot usage. This article organizes the specifications, added metrics, reporting periods, and permission requirements based on the official changelog.
Through this article, we briefly outline the prerequisite knowledge for organizational and enterprise administrators to safely and practically understand Copilot CLI customization usage, along with points to keep in mind when using the API.
1. Purpose and scope of this article
The purpose of this article is to accurately interpret the components of the added features and reporting specifications based on the official GitHub changelog titled 'Agentic CLI customizations now in the usage metrics API'.
Since this has 'not been verified on actual hardware', this article does not cover actual API requests or retrieving environment-specific output results using PowerShell, but focuses on organizing the official specification descriptions.
2. Prerequisites and notes
The primary sources state that checking custom CLI usage status via the Usage Metrics API requires the following roles, permissions, and policy enablement.
Target roles: Enterprise owners, billing managers, organization owners, or users with
View Copilot Metricscustom organization or enterprise roles granted permissions.Prerequisites: The Copilot usage metrics policy must be enabled.
Privacy protection specifications: For privacy protection, user-defined names are not displayed; only recognized GitHub-provided item names are shown.
Skills, custom agents, MCP servers, and plugins are grouped under the label
other.For custom slash commands, the label
customis used, consistent with the behavior of traditional Copilot CLI telemetry.
3. Types and Specifications of Added Metrics
According to official information, this update adds activity metrics for skills, custom agents, Model Context Protocol (MCP) servers, slash commands, and plugins.
These fields appear in enterprise and organization "per-user and aggregate daily reports," "per-user 28-day reports," and in the "entries" of day_totals aggregate 28-day reports.
The reports address the following two main questions:
Which items are used the most?
How many different types of items were used?
4. Array Fields for Identifying Frequently Used Items
To understand frequently used items, the following array fields are provided. These list the top five items with the highest activity records, each containing interaction_count.
totals_by_skilltotals_by_custom_agenttotals_by_mcptotals_by_slash_cmdtotals_by_plugin
Interaction Count Specifications
Depending on the category, interaction_count defines what is counted.
Skills, slash commands, and plugin skills: Counts the number of invocations.
Custom agents: Counts the number of starts.
MCP servers: Counts the number of times the Copilot CLI attempts to connect or reconnect to the server. Both successful and failed attempts are counted. Invoking multiple tools from the same connected server does not increase this count.
Plugins: Counts only the invocation of skills associated with the plugin. All plugin interactions also appear in the skill totals, but skill interactions not originating from the plugin appear only in the skill totals. Because plugin totals are a subset of skill totals, the documentation explains that the two should not be simply added together.
5. Count fields for measuring diversity
To understand the diversity of items utilized, the following fields are provided to count the number of distinct items used. These counts include items outside the top five, and comparing them over time allows you to verify whether usage diversity is expanding.
distinct_skill_use_countdistinct_custom_agent_use_countdistinct_mcp_use_countdistinct_slash_cmd_use_countdistinct_plugin_use_count
Aggregation unit for counts
Per-user report: Each individual item used by that user is counted once.
Aggregate report: Each individual item used by anyone within the enterprise or organization is counted once overall (not per user).
6. Behavior when data is absent and important notes
The primary documentation also notes the behavior when no activity exists.
If there is no matching activity, the array is empty and the count is zero.
If customization data is unavailable, the field is
nullor absent.
7. Summary
We have organized the official information from the GitHub Changelog titled "Agentic CLI customizations now in the usage metrics API". The points to check and constraints before execution are as follows.
Points to check prior to execution:
Whether the Copilot usage metrics policy is enabled in the utilizing organization.
Whether the account accessing the API includes
View Copilot Metricspermissions.
Constraints and considerations:
To protect privacy, user-defined names are not displayed and are grouped as
otherorcustom.Because plugin metrics are a subset of skill metrics, they should not be simply summed up.
The number of MCP server interactions represents connection attempts rather than tool calls.
Because this is [prior to actual device verification], the actual API response JSON structure and acquisition results in PowerShell are not shown. For detailed schema verification, refer to the official Copilot usage metrics API documentation.
References
Update history of this article
This article has been reviewed and necessary corrections have been reflected through an automated review and update workflow utilizing generative AI.
September 26, 2026
- ChangesFixed a missing character at the beginning of Chapter 7.
