Hello Cursor Support team,
I’m currently evaluating how to operationalize internal cost reporting and chargeback for Cursor Teams usage within my organization, and during this work I identified discrepancies between cost figures coming from different Cursor sources.
I would appreciate your clarification on how these sources should be interpreted and which one should be treated as authoritative for financial reporting and reconciliation.
What I’m seeing
For the same team and the same UTC day, I compared usage cost data from three sources:
-
Dashboard UI
-
Dashboard CSV export
-
Admin API endpoint: /teams/filtered-usage-events
In my testing, the number of events and token totals can match between the dashboard export and the API dataset for the same UTC period, but the total cost still differs.
Example
For one of the analyzed UTC days, I observed approximately the following:
-
Dashboard UI total: about $3,354.88
-
Dashboard CSV export total: about $3,331.97
-
/teams/filtered-usage-events aggregated charged cost total: about $3,377.78
In other scenarios, even when:
-
the event count matched,
-
timestamps matched,
-
and token totals matched exactly,
the summed event-level cost from the API still didn’t exactly match the dashboard CSV export total.
What I already validated
Based on my analysis, the discrepancy doesn’t appear to be caused simply by:
-
missing events,
-
incorrect pagination handling,
-
timezone filtering mistakes,
-
or token count mismatches.
I also checked the event-level fields returned by /teams/filtered-usage-events, including:
-
tokenUsage.totalCents
-
cursorTokenFee
-
chargedCents
From the documentation, my understanding is that:
-
chargedCents = total amount charged for the event
-
chargedCents should be used to reconcile event-level costs with /teams/spend totals
However, in practice I still see differences between:
-
Dashboard UI totals
-
Dashboard CSV export totals
-
SUM(chargedCents) / 100 from /teams/filtered-usage-events
Questions
Could you please clarify the following:
-
Why can the cost totals differ between:
-
Dashboard UI
-
Dashboard CSV export
-
/teams/filtered-usage-events API
-
-
Are these three sources expected to match exactly for the same team and same period, or are there known differences in calculation / presentation logic?
-
Does Dashboard UI or Dashboard CSV apply any additional logic such as:
-
rounding at a different aggregation level,
-
different treatment of Cursor Token Fee,
-
included usage vs on-demand presentation logic,
-
model remapping/grouping (for example Auto / Other / internal buckets),
-
adjustments, refunds, credits, or other billing-layer corrections?
-
-
When there is a discrepancy, which source should be treated as authoritative for financial reconciliation and internal reporting:
-
Dashboard UI
-
Dashboard CSV export
-
/teams/filtered-usage-events
-
-
If /teams/filtered-usage-events isn’t expected to match dashboard totals exactly, what is the correct API-based method to reproduce the same totals that the dashboard displays?
-
If Dashboard CSV export is generated using a different internal aggregation strategy than the public Admin API, is there any documented or recommended way to reproduce dashboard-equivalent totals programmatically?
-
Are there any known differences in how costs are represented for:
-
Teams vs Enterprise
-
strategy=tokens vs strategy=spends in dashboard exports
-
included usage vs on-demand usage
-
request-based vs token-based billing plans
-
My goal
My goal is to build a reliable internal reporting and chargeback process based on Cursor Admin API data. To do that correctly, I need to understand:
-
why these discrepancies occur,
-
what each source is intended to represent,
-
and which source should be trusted as the canonical source for billing reconciliation.