GitHub Copilot usage metrics API now reports usage by client
A new totals_by_client field separates Copilot requests from OpenCode, Claude Code, Codex and other non-IDE tools that were previously missing from reports.
GitHub’s Copilot usage metrics API now reports which tool actually sent each request, not just which IDE. A new totals_by_client array landed on 23 October in the enterprise, organisation, enterprise-user and organisation-user reports, across both the 1-day and 28-day windows.
Each entry names the client, one of opencode, claude_code, codex, zed, copilot_cli, github_app, cloud_agent, code_review, vscode or jetbrains, alongside request count, session count and token usage split by prompt and output. Until now, anything using Copilot as a model provider without sending IDE telemetry, such as a coding agent running outside an editor, simply didn’t appear in the existing totals_by_ide breakdown.
Why it matters: teams paying for Copilot seats while also running agents like OpenCode or Claude Code against the same models had no way to see that usage in GitHub’s own reporting. Counting by client on GitHub’s side closes that gap, so a usage or cost review no longer silently undercounts anything that isn’t a traditional editor plugin.