Reports

If the dashboard answers "how much are we spending," Reports answers "show me exactly where that came from." It's the detailed, line-item view behind every number on the dashboard.

How it's organized

Everything sits in one view, in three parts:

• Overview: the date range, platform filter, and export at the top, with your total spend and number of platforms for the period.
• Breakdown: a ranked list of spend by platform and by model, the "where did the money go" answer without digging any further. Long lists scroll, so nothing is ever grouped away.
• Records: a table of individual cost records, newest first, paged so it stays fast no matter how much history your account has.

Each record shows the date, platform, model, API key, input and output tokens, and cost. Token and model columns show a dash when they don't apply to a record. On Business and Scale there's also a Budget column showing which budget each record counts toward, and a vs list column: what the provider billed compared with the same tokens at its published prices (cached input, cache writes and long-context rates included). It's shown for provider-billed records only, since an estimate is list price by definition. On Free, the model and API key columns show a dash: breakdowns by model and key are part of Pro and above.

Where the numbers come from

Every platform's daily total, and the cost of each model, are the provider's own billed figures. A cost marked ≈ is one TokenFront worked out from them:

• Split across API keys. OpenAI and Anthropic report cost per model but not per API key, so TokenFront divides each model's cost across your keys by what each key used. The split always adds back up to the provider's exact figure.
• Not billed yet. AWS posts Bedrock charges about a day late. Until it does, the most recent day is estimated from the tokens used at list price, then replaced with the billed amount on a later sync.

Filtering what you see

• Date range: the same timeline filter used on the dashboard (7D, 30D, month-to-date, a custom range, and so on).
• Platform: narrow the view to one connected platform, or leave it on "All platforms."
• Search: type a model name or API key to filter the detailed records down to just those matching.
• Sort: newest, oldest, highest cost, or lowest cost.

The totals at the top (Total, Platforms) and the ranked breakdown always reflect your full selected date range. Searching or sorting the records table narrows what's listed there without changing those totals, so the top-line numbers you'd report to someone else never shift out from under you while you're digging through individual records.

If you use the same combination of filters often, save it as a saved view from the records table's filter bar. It'll be there next time without re-entering everything. Free includes one saved view; paid plans have no limit.

Exporting

On Pro and above, click Export CSV to download the records for your current date range, platform filter, and search as a spreadsheet (on Business and Scale it includes the budget column and a list_price_usd column). When a search is active, a note beside the button says how many filtered rows you're exporting. Exports are capped at 50,000 rows, which covers the overwhelming majority of accounts comfortably. If a given export would exceed that, you'll see a warning telling you how many rows actually matched before you download, so you know up front whether you're getting everything or need to narrow the date range first.

Arriving here from an Intelligence link

Anomalies and Insights sometimes link straight to Reports pre-filtered to the specific record they're about. You'll land here with the search box already filled in, so you can go straight to the relevant rows instead of hunting for them.