On October 23, GitHub added a new totals_by_client breakdown to its Copilot usage metrics API, so administrators can now see model traffic from tools outside the editor, including OpenCode, Claude Code and Codex. According to GitHub’s changelog, the change closes a gap in which usage from those tools went missing because they do not send editor telemetry. The breakdown appears in enterprise, organization, enterprise-user and organization-user reports, in both 1-day and 28-day forms. It counts requests and tokens by client, not prompts, and GitHub says the grouping is best-effort, based on the integration ID each tool sends.
What the new breakdown counts
Each totals_by_client entry represents one client tool. GitHub says an entry carries a request count, a session count and token usage, covering prompt tokens, output tokens and the average tokens per request. It also carries a totals_by_model breakdown, so the same traffic can be viewed by model as well as by tool.
The model breakdown records how each model was selected. GitHub lists three modes: user_selected, for a model the user requested directly; auto, for a model chosen automatically; and product_selected, for a model chosen by the product itself. Modes that are not reported individually appear as other, and values that cannot be normalised appear as unknown.
In the aggregated 28-day reports, GitHub says totals_by_client appears inside each day_totals entry. The change is additive: existing fields are unchanged, and the array is empty when there is no usage to report for a client.
Which tools are named
GitHub names ten clients in the announcement: opencode, claude_code, codex, zed, copilot_cli, github_app, cloud_agent, code_review, vscode and jetbrains. Tools that GitHub does not recognise are grouped under other, rather than listed individually.
The inclusion of OpenCode is the headline case, because GitHub says OpenCode usage was not previously captured in these reports. Claude Code and Codex matter for the same reason: both are agent tools that developers run outside the editor, and their traffic previously sat outside the reporting that administrators relied on.
Access, limits and what it does not measure
The reports are available to enterprise owners and billing managers, organization owners, and anyone holding a custom organization or enterprise role that grants the View Copilot Metrics permission. GitHub also says the Copilot usage metrics policy must be enabled before the data appears.
Two limits deserve attention. First, request_count counts model calls, not prompts, and GitHub warns that one prompt can trigger several model calls, so the figures should not be compared directly with prompt or interaction counts. Second, clients are identified by the integration ID they send with each request, which GitHub describes as a best-effort grouping, not a verified identity.
GitHub draws a further distinction. Totals_by_client measures model traffic, meaning requests and tokens, from any tool that uses Copilot. That differs from agent app activity, which counts jobs started by agent apps such as Claude and Codex on GitHub. The two measures answer different questions, and GitHub says they should not be confused.
There is one reporting quirk for two clients. Requests from Copilot code review and Copilot cloud agent report product_selected, because the product chooses the model in those cases. Code review models are reported as unknown.
Why It Matters
GitHub Copilot’s usage reports previously told administrators most clearly about work inside the editor, while a growing share of Copilot-powered work happens elsewhere. Developers now run Copilot as the model provider behind separate agent tools, and GitHub says those tools often do not send editor telemetry, which left their usage missing from totals_by_ide and hard to account for. Counting requests on the server closes that gap.
For an organization, the practical gain is a complete picture of where its Copilot traffic goes. Administrators can compare request and token volume across tools, see which outside tools drive usage, and, in GitHub’s words, base tool support and governance decisions on real usage rather than on editor telemetry alone. A team that standardised on an outside agent can now show that choice in its own reports, and a team weighing whether to support a new tool can measure its existing footprint first.
Background
GitHub has spent October tightening Copilot’s administrative controls, including a global default policy for unconfigured Copilot features and a consolidated GitHub app for Microsoft Teams that turns notifications into code. The usage metrics change follows the same pattern: it gives administrators clearer visibility over a Copilot deployment that now stretches well beyond the editor.
Source: GitHub Changelog
0 Comments