Field Guide: How to Audit GitHub Copilot Usage After the July 28 Metrics Rollout

GitHub changed Copilot reporting five times between July 17 and July 28, 2026. If you are still checking one dashboard and calling it done, you are missing the new signals.

What actually changed in July

On July 17, GitHub added GitHub Copilot app activity to the usage metrics API. That gave admins a first clean read on app sessions, prompts, requests, and token usage at the organization and enterprise level.

Also on July 17, GitHub shipped repository-level usage metrics. The new repos-1-day reports break out pull request activity for Copilot coding agent and Copilot code review per repository, which is much closer to how engineering leaders actually ask where value is showing up.

On July 20, GitHub updated the user usage page so Copilot Business and Enterprise users can see AI credits consumed in the current billing cycle even when no individual budget is set. The same day, cost center AI credit pools moved into the billing UI, so finance and platform teams no longer need the API for that basic control.

On July 22, GitHub added the Copilot usage metrics impact dashboard. This is the opinionated layer on top of the raw data: adoption phases, throughput, merge velocity, and suggested next steps for moving users deeper into agent workflows.

On July 28, GitHub folded Copilot app activity into the broader rollups. That is the important follow-through. After this change, app usage stopped being a side counter and started appearing in user, feature, model, language, and lines-of-code views through fields like used_copilot_app and copilot_app.

The fastest way to read the new stack

If you are an individual contributor, start with the built-in usage view. GitHub now shows usage this cycle in Copilot settings, and individual plan users can also inspect the AI usage page under billing. That answers the simple question: am I burning credits faster than I thought?

If you run an organization, open the impact dashboard before you touch the API. The dashboard is the shortest path to finding three buckets: licensed but passive users, active code-first users, and people already leaning into agent-first or multi-agent workflows.

If you need proof by repository, then pull the repository report for a single day. The report only includes repos that had activity, which keeps it useful. You are looking for hotspots: repositories where Copilot coding agent is opening or merging pull requests, and repositories where code review suggestions are actually happening.

If you own budget controls, finish in billing. July 20 matters because AI credit pools for cost centers are now manageable in the UI, and GitHub budget controls still determine whether usage is merely metered, warned, or blocked.

One report worth automating first

Do not start with a giant warehouse pipeline. Start with one daily repository report and one 28-day user report.

bash

curl -L \
  -H "Accept: application/vnd.github+json" \
  -H "Authorization: Bearer $GITHUB_TOKEN" \
  -H "X-GitHub-Api-Version: 2026-03-10" \
  "https://api.github.com/orgs/ORG/copilot/metrics/reports/repos-1-day?day=2026-08-13"

That gives you a lazy but defensible operating loop:

Impact dashboard for the story
users-28-day for adoption depth and app activity
repos-1-day for where pull request work is actually landing
Billing and budgets for whether the spend shape is acceptable

What to stop using

The old Copilot Billing Preview app was retired on August 3, 2026. GitHub is clearly steering users toward the AI usage page, budgets, cost centers, and the usage metrics API. If your internal runbook still points people to the preview app, update it now.

Bottom line

The July 28 rollout is the hinge. Before it, Copilot app usage was visible but isolated. After it, the app became comparable with the rest of the Copilot surfaces in the same reporting system. That makes July 2026 the first time GitHub Copilot reporting feels usable for both engineering leadership and finance, without a pile of custom stitching.