GitHub updated its Copilot usage metrics API on October 22, 2026 so organization-created skills, custom agents, Model Context Protocol (MCP) servers, slash commands, and plugins can be reported by their configured names rather than being collapsed into generic other or custom buckets. For QA teams that are rolling out internal AI-assisted test tools, that adds a more useful adoption signal.
What changed in Copilot usage metrics
GitHub says the named entries are available in enterprise and organization reports, including per-user one-day and 28-day reports, aggregated one-day reports, and day_totals in aggregated 28-day reports. Each totals_by_* list can now contain up to 25 entries per record, up from five; remaining activity is grouped into a trailing other entry.
The update does not change how interaction_count or the distinct-use fields are calculated. GitHub also notes that names begin appearing only for report days processed after the release, so historical data will still contain generic buckets.
Why this matters for QA engineers
Teams increasingly expose test-data helpers, browser-control tools, defect-triage agents, and release-check skills through AI coding workflows. A named metric is not proof that a tool produced a correct result, but it can show whether the intended workflow is actually being adopted. That gives QA leads a better starting point for deciding where to collect test evidence, run a usability review, or retire an unused integration.
- Measure adoption by workflow: define a stable, non-sensitive name for each approved test skill or MCP server before rollout.
- Pair usage with quality evidence: compare adoption trends with CI pass rates, escaped-defect signals, flaky-test rates, and human review outcomes.
- Test reporting coverage: include the supported Copilot clients in a release checklist and record their versions.
- Protect report hygiene: review configured names before exporting metrics; GitHub says names resembling file paths remain grouped and names are trimmed to 128 characters.
A practical validation checklist
- Create a small pilot with one custom QA skill, one agent, and one MCP-backed tool.
- Run representative interactive and automated workflows from each supported client.
- After a reporting day is processed, verify that configured names appear in the expected metric arrays and that values are plausible.
- Confirm that a blocked-telemetry or outdated-client scenario is treated as an observability gap, not as evidence of zero usage.
- Keep raw test results and pull-request review as the quality gate; the metrics describe activity and adoption, not correctness.
Coverage includes Copilot CLI interactive sessions and prompt mode, the GitHub Copilot app, Agent Client Protocol integrations such as JetBrains, and the Copilot SDK, subject to the versions GitHub documents. GitHub also cautions that the metrics come from client telemetry and may undercount when clients are outdated or telemetry is blocked. Access requires the relevant Copilot metrics permission and the usage-metrics policy to be enabled.