Copilot metrics now name internal skills, agents and MCP servers

Configured names make internal customizations visible

GitHub’s October 22 Copilot update adds configured names for customer‑defined skills, custom agents, Model Context Protocol (MCP) servers, slash commands and plugins to usage reports. Activity that previously appeared under other or custom can now carry the name of the internal customization being used.

The affected arrays are totals_by_skill, totals_by_custom_agent, totals_by_mcp, totals_by_slash_cmd and totals_by_plugin. The change covers enterprise and organization reports: per‑user 1‑day and 28‑day reports, aggregated 1‑day reports, and day_totals within aggregated 28‑day reports. For an administrator rolling out an internal agent or skill, these labels make it easier to locate its reported use.

Lists expand from five entries to as many as 25

Each totals_by_* list now contains up to 25 entries per record, increased from five. Remaining activity goes into a trailing other entry, according to GitHub’s changelog. That exposes more of the customization mix while retaining a combined bucket for activity beyond the displayed entries.

GitHub says the calculations for interaction_count and the distinct_*_use_count fields are unchanged. A more specific label or a longer list therefore supplies additional detail about the report; it does not, by itself, establish that developers are using Copilot more or producing better results.

Client and SDK versions determine name coverage

GitHub lists coverage across Copilot CLI interactive sessions and prompt mode, the GitHub Copilot app, Agent Client Protocol integrations such as JetBrains, and the Copilot SDK. Prompt mode, the app and ACP integrations require Copilot CLI 1.0.97 or later. The Copilot SDK requires version 1.0.21 or later.

IDE coverage depends on which experience produced the telemetry. Agent experiences running on Copilot SDK 1.0.21 or later can report names; an IDE’s built‑in chat telemetry does not report them. An administrator investigating a missing name should first check the originating experience and its version, rather than assume the customization was unused.

Names need review before reports leave the organization

Configured names are trimmed and limited to 128 characters. Names that resemble file paths remain grouped under other or custom. GitHub advises reviewing names before sharing reports outside the organization.

The useful administrative label may also disclose an internal project or tool name. Reports should be reviewed with that naming context in mind; the path‑like‑name rule should not be treated as a guarantee that every sensitive label has been removed.

Enable access and preserve the reporting limits

Reports require the Copilot usage metrics policy to be enabled. GitHub says access is available to enterprise owners and billing managers, organization owners, and custom organization or enterprise roles granting View Copilot Metrics.

Names begin appearing with report days processed after the release. Earlier days keep other or custom, so historical reports do not gain the same named breakdown. The data also comes from client telemetry and can undercount activity when clients are outdated or telemetry is blocked. These reports support an adoption readout, with those limits retained, rather than exact accounting of all work.

OpenTools’ separate coverage of external‑client metrics explains request and token traffic grouped by the client using Copilot. The two views answer different questions: one locates internal customizations in covered experiences; the other groups model traffic by client. Neither establishes productivity, code quality or conversion outcomes on its own.

A letterbox for GitHub’s office in Spuistraat, Amsterdam, photographed on November 9, 2023. Archive company context. Photo: Rrustema, via Wikimedia Commons, CC0 1.0 Universal. Resized and cropped for display.

Share this article

Related News

Vercel explains how agent plans select production permissions

AWS Continuum brings vulnerability fixes to GA, with approval before autonomous mode

NVIDIA Agent Knowledge Foundation connects company data with agent controls

Copilot metrics now name internal skills, agents and MCP servers

Configured names make internal customizations visible

Lists expand from five entries to as many as 25

Client and SDK versions determine name coverage

Names need review before reports leave the organization

Enable access and preserve the reporting limits

Sources

Tags

Copilot metrics now name internal skills, agents and MCP servers

Configured names make internal customizations visible

Lists expand from five entries to as many as 25

Client and SDK versions determine name coverage

Names need review before reports leave the organization

Enable access and preserve the reporting limits

Sources

Tags