GitHub Adds Per-User AI Credit Usage to the Copilot Metrics API
GitHub's Copilot usage metrics API now includes ai_credits_used in user-level reports, giving administrators a clearer signal for adoption and budget reviews.
GitHub added an ai_credits_used field to user-level Copilot usage metrics reports. The changelog says the value is available in single-day and 28-day reports at enterprise and organization levels, and that it is derived from the same AI-credit consumption data used by the usage-based billing API. GitHub also warns that the field is a consumption signal, not an invoice total.
Key takeaways
- The new
ai_credits_usedfield reports total AI credits consumed per user across Copilot activity. - GitHub says the field appears in
users-1-dayandusers-28-dayuser-level reports for enterprises and organizations. - The value is not broken down by feature, model, or surface, so it cannot answer which Copilot feature created the spend.
- The metric is meant for adoption and consumption analysis; billing pages remain the source for invoicing.
- Admins can now put usage, adoption, and budget review closer together without scraping separate views.
Practical LinkLoot angle
This is a FinOps upgrade for teams rolling out coding agents at scale. It helps admins spot outlier consumption, compare teams, and decide where training or policy changes are needed before usage-based costs become a surprise.
| Metric source | Best use | Limitation | Source |
|---|---|---|---|
ai_credits_used in Copilot usage metrics | Per-user consumption trend analysis | Not feature-, model-, or surface-specific | GitHub Changelog |
| Copilot usage metrics API | Adoption and engagement reporting | Requires proper admin/API permissions | GitHub Changelog |
| Billing views and usage-based billing API | Invoice and budget reconciliation | Less useful for day-to-day enablement coaching | GitHub billing docs |
The practical workflow is simple: pull the 28-day user report, join it with team data where available, flag unusually high credit use, then compare it with active-user and feature-adoption metrics. Do not treat the number as a chargeback amount unless finance has reconciled it against billing data.
What to verify before you act
Confirm your token permissions before building dashboards around the endpoint. GitHub's docs note Copilot usage metrics endpoints require the relevant Copilot metrics permissions, and some reports are scoped by organization or enterprise. Also verify how your company wants to handle per-user productivity metrics, because credit consumption can become sensitive employee analytics if published too broadly.
For teams comparing agent tools and operating costs, LinkLoot's AI tools guide is a useful companion: /guides/ai-agent-tools.
It is a user-level Copilot usage metric showing total AI credits consumed by a user across Copilot activity.