GitHub Copilot usage reports now show names of custom skills, agents and MCP servers

From pooled activity to named adoption

GitHub’s announcement covers organization-created skills, custom agents, MCP servers, slash commands and plugins in its usage metrics API, the interface organizations use to retrieve activity reports. MCP, or Model Context Protocol, connects AI software to external tools and data.

According to the announcement, each report record carries lists of totals broken out by item type, such as skills or MCP servers. Configured names appear in enterprise and organization reports: per-user daily and 28-day reports, aggregate daily reports, and daily totals within aggregate 28-day reports.

Previously, customer-defined items appeared under generic labels such as other or custom, GitHub says. Named entries let administrators identify adoption of a specific internal customization rather than follow only a pooled category. Activity beyond the displayed entries goes into a trailing other entry, an overflow bucket.

The announcement limits reported names to 128 characters and trims them; names resembling file paths remain under generic labels. Named reporting begins with report days processed after release, while earlier days retain their generic labels. That leaves older reports without the names needed to follow the same customization backward through its history.

Which clients report names

GitHub lists coverage across interactive sessions in Copilot CLI, its command-line interface, alongside prompt mode, the Copilot app, ACP integrations such as JetBrains, and the Copilot software development kit, or SDK.

The announcement specifies CLI version 1.0.97 or later for prompt mode, the app and ACP integrations. SDK-based reporting requires version 1.0.21 or later. Agent experiences in integrated development environments, or IDEs, can report names when running on that SDK version, but an IDE’s built-in chat telemetry does not report customization names.

Outdated clients and blocked telemetry can undercount activity, according to GitHub, which recommends these metrics for adoption trends rather than exact accounting.

Existing counters keep their limits

GitHub introduced CLI customization metrics on September 17. Its October update changes labels and list capacity while leaving interaction_count and distinct_*_use_count, the per-item interaction and distinct-use counters, unchanged.

The September announcement and General Analysis’s subsequent guide explain that MCP interaction counts measure connection and reconnection attempts, including failures. Repeated tool calls over an existing connection do not increase that counter. A newly visible server name identifies the connection activity being counted without supplying a tally of executed tools.

Those sources also explain that plugin totals are a subset of skill totals, so adding the two would double-count activity. Aggregate distinct-use counts count each customization once across an organization or enterprise, rather than once per user; per-user reports provide the separate user-level view.

Access requires an enabled usage metrics policy and an eligible administrative role or a custom role granting View Copilot Metrics, GitHub says. Its REST documentation specifies the enterprise policy setting Enabled everywhere for enterprise endpoints. The announcement also advises reviewing configured names before sharing reports externally.

Not yet confirmed

• The October named-reporting change is supported only by GitHub’s announcement. The available independent guide predates it, does not validate named reporting in an enterprise tenant, and provides no second independent corroboration.
• GitHub has not specified tenant-by-tenant rollout completion or the first named report available to each tenant.
• The announcement does not clarify whether the trailing other entry counts toward the 25-entry maximum.

What to watch

• The first processed reports containing configured names and their coverage across client versions.

Related coverage

• AIGitHub Actions retires macOS 14 runners on November 2, with brownouts through October 30
• AIGitHub makes agentic autofix generally available for code scanning alerts
• AIGitHub secret scanning adds detectors for Lovable, Supabase and Pydantic tokens