A GitHub Copilot usage dashboard connects Code Editor, OpenCode, Copilot CLI, and Cloud Agent activity to reports on models and tokens.
GitHub Copilot’s usage reports now account for more activity outside the IDE, including OpenCode. Announced October 23, 2026, the new totals_by_client array captures server-side model traffic previously missed by editor-based reporting. It covers enterprise, organization, enterprise-user, and organization-user reports in both 1-day and 28-day formats.

A GitHub Copilot usage dashboard connects Code Editor, OpenCode, Copilot CLI, and Cloud Agent activity to reports on models and tokens. What client-level reporting reveals​

Each client entry contains successful request counts, distinct sessions, token usage, and a model breakdown. Recognized labels include opencode, claude_code, codex, zed, copilot_cli, github_app, cloud_agent, code_review, vscode, and jetbrains; unrecognized clients become other.

The model breakdown distinguishes direct user selection (user_selected), automatic selection (auto), and product selection (product_selected). Remaining selection modes use other or unknown. Code review and cloud agent requests use product_selected; code review models appear as unknown.

Requests are not prompts: one prompt can generate several model calls. Nor are these agent-app job counts. Client labels come from request integration IDs, so GitHub describes them as best-effort groupings—not verified identities. Existing fields remain unchanged, and no reportable usage produces an empty array.

How administrators can use the reports​

GitHub’s REST documentation establishes an important implementation detail: report endpoints return download links, rather than the complete metrics dataset directly. Reports are generated daily, and the signed download URLs expire. The latest 28-day report ends on the most recent fully processed reporting day—not necessarily today.

For an enterprise reporting workflow:

  1. Check policy and authorization. GitHub’s API documentation requires the Copilot usage metrics policy to be enabled everywhere for the enterprise. Enterprise owners, billing managers, and appropriately authorized metrics readers can retrieve enterprise reports.
  2. Select the reporting window. The daily endpoint is GET /enterprises/{enterprise}/copilot/metrics/reports/enterprise-1-day, with a required day parameter in YYYY-MM-DD format. The rolling endpoint is GET /enterprises/{enterprise}/copilot/metrics/reports/enterprise-28-day/latest.
  3. Download the returned files promptly. Process the report files referenced by download_links, retaining the returned reporting dates alongside the data. GitHub’s response examples identify NDJSON report files.

For aggregated 28-day reports, inspect totals_by_client inside each day_totals entry. Organization owners and custom roles with View Copilot Metrics permission also have access within the applicable scope.

Better visibility, not a productivity score​

The practical recommendation is to chart client traffic separately from prompt interactions and agent jobs. Otherwise, a dashboard risks comparing unlike units and presenting the result with unwarranted confidence. That recommendation follows directly from GitHub’s distinction between model calls, prompts, and job starts.

For IT teams deciding which developer tools to support, this is a useful accounting improvement. It makes previously overlooked Copilot traffic visible—but a busier model is not, by itself, evidence of a more productive developer. Treat the figures as operational signals, not a leaderboard.

 

References

  1. Usage metrics API adds usage from tools like OpenCode GitHub Changelog 2026-10-23T23:42:03+00:00
  2. REST API endpoints for Copilot usage metrics - GitHub Docs docs.github.com