A more useful inventory of internal tools
GitHub announced on October 22 that Copilot usage reports can identify an organization’s own skills, agents, MCP servers, commands and plugins by their configured names. Previously, custom items could disappear into general categories, making it harder to tell which internal tools people were using.
The accompanying reference documents up to 25 named entries in each applicable daily breakdown, with remaining activity combined. An item missing from that list therefore cannot be assumed unused. The reports measure recorded events; they do not establish how many people benefited from a tool.
Check the collection path first
Access requires the Copilot usage-metrics policy and an authorized reporting role or permission. Client versions also matter: GitHub specifies CLI 1.0.97 for prompt mode, the Copilot app and ACP integrations, and SDK 1.0.21 for SDK-based collection. Built-in IDE chat does not supply these customization names.
Names appear in newly processed report days; earlier reports are not automatically rewritten. GitHub warns that disabled telemetry or older clients can leave gaps. The reviewed material does not establish a separate price or country-specific rollout for this change. CGH has not tested the reporting service.
CGH perspective
Before comparing tools, record which clients and versions contributed data. Otherwise, an apparent increase may reflect better collection rather than greater adoption. Start with a small set of known internal tools and compare report entries with an observed sample of activity.
Review configured names before circulating exports, because internal project labels may reveal information beyond the intended audience. Use the results to decide where training or investigation would help, then assess completed work separately. More recorded activity alone does not demonstrate better output.