🤖 GitHub Copilot AI Usage Viewer

A browser-based visualizer for the GitHub Copilot AI Usage Report CSV. Drop in your billing export to explore spending by org, member, model, and day.

CSV data is processed in your browser and is not uploaded.

Use in browser: https://daichiueura.github.io/copilot-ai-usage-viewer/ (Demo)

Usage

1. Go to GitHub → Billing → AI usage → Get usage report and download the CSV.
2. Open index.html in your browser.
3. Drop the CSV onto the page.

Open a CSV from URL

Use csv= to load a CSV by URL. Relative URLs work when the CSV is hosted on the same site:

https://daichiueura.github.io/copilot-ai-usage-viewer/?csv=reports/ai-usage-report.csv&tab=overview

External URLs are supported when the CSV server allows cross-origin requests:

https://daichiueura.github.io/copilot-ai-usage-viewer/?csv=https://example.com/ai-usage-report.csv&tab=overview

The resolved URL must be HTTP(S), and CSV files are limited to 10 MB.

A second CSV can be opened alongside the first with users_csv= — see Automation.

When the host dates the response with a Last-Modified header, each panel's caption adds when its copy was published, in your own time zone. A feed republished on a schedule can stall without the rows ever looking wrong, and that date is what shows it.

Measure against a budget

Use net_limit= to pass a budget. It draws a horizontal reference line on the Overview cumulative chart, against the cumulative Net (metered) line, and adds a header stat showing how much of the budget the billed amount has used. That share covers the whole file, over whatever range it holds — the range each panel's caption names — so pass a budget for the same period. The value is supplied here, not read from the CSV. With no net_limit, neither is shown.

https://daichiueura.github.io/copilot-ai-usage-viewer/?csv=reports/ai-usage-report.csv&net_limit=10000

Views

• Overview — cumulative spend, daily total, model share; metered billing overlay when applicable
• Members — per-member ranking by Gross and by Net, day-by-day usage for the top members, usage by member and model, and a sortable detail table

A tab is offered only when the loaded CSV can answer it, and the viewer notes why when one is missing. A full export answers both.

How usage is interpreted

The viewer aggregates usage from the standard billing columns (gross_amount / net_amount / quantity). The aic_quantity / aic_gross_amount preview columns were deprecated and zeroed by GitHub on 2026-06-01, so they are ignored.

For the CSV interpretation policy used by this viewer, see docs/csv-interpretation-policy.md.

Supports EN / 日本語. Validates the CSV format on load.

Automation (optional)

GitHub only shows AI credit usage to billing admins. To share it with people who can't see it in the GitHub UI, the automation/ folder has small, dependency-free Node scripts that rebuild the data from the API and write CSVs this viewer can open. There are two recipes:

• Organization totals — gross and net from the AI credit usage report API. No per-member rows, so the result is safe to publish to non-admins.
• Per-user consumption — how much each member consumed, day by day, from the Copilot usage metrics reports. Gross only: no API attributes net to an individual member, so those columns are left empty rather than estimated.

Run them on demand or on a schedule (e.g. GitHub Actions), then host the CSV for your team. See automation/README.md for the scripts, tokens, and what each CSV does and does not carry.

Open both CSVs together

Neither recipe answers every question, so each one alone shows only the tabs it can fill: the organization totals have no member dimension and show the Overview only, while the per-user rows carry no billed amount and show Members only, reporting Net as not available rather than as fully covered. Add users_csv= to open both, and each tab reads the one that can answer it:

https://daichiueura.github.io/copilot-ai-usage-viewer/?csv=reports/ai-usage.csv&users_csv=reports/ai-usage-by-user.csv

Which CSV drives which tab follows from its content, not from the parameter name, so dropping both files on the page works the same way. Each panel names the file it is reading, the days that file covers, and whether its amounts include Net — the two feeds have independent timelines, and neither is trimmed to match the other.

Testing

Two suites: viewer e2e (Playwright drives index.html and loads CSV fixtures) and automation unit tests (node --test, covering the companion scripts). npm test runs both.

npm install
npx playwright install-deps chromium
npx playwright install chromium
npm test          # or: npm run test:e2e  /  npm run test:unit