Rate limits
Limits count in 5-minute windows. A request over a limit gets 429rate_limited, which is retryable: the SDK throws RollupError with code rateLimited, and retryAfter when the server says how long to wait. Wait and try again.
Per device
A
query or profile usually takes a few steps.
Per app
These are shared by every install of your app.
Rollup sets these per app. Your app’s page in the dashboard shows its current limits. If your app needs more, ask Rollup to raise them.
Monthly token budget
Each app has a budget of model tokens per month: the input and output tokens of everyquery and profile step, across all installs. The default is 5,000,000 tokens per calendar month in UTC.
Before each step, Rollup checks the tokens your app has used this month. Once they reach the budget, query and profile get 429 app_quota_exceeded, which is not retryable, with a message naming the date the budget resets: the first day of the next month, UTC. A step that is already running finishes, so the total can go slightly over the budget.
Search, indexing, place data and everything else keep working. Your app’s page shows this month’s tokens against the budget, and Rollup can raise the budget.
Usage
Rollup counts your app’s usage per day in UTC and shows it on your app’s page:- devices registered,
- place data and geography requests,
querysteps andprofilesteps,- input tokens and output tokens,
- searches, as reported in the SDK’s run reports.

