Usage analytics
Every request measured
Usage shows what you spend, how fast requests come back and how often they fail. Use it to decide which models and providers to keep.
What Usage shows
Spend by source
Spend split by model, provider, key and teammate, beside the period before it.
Speed
The p95 latency of your served requests.
Reliability
How many requests were served, failed at the provider or were refused by your limits.
Cache hits
The share of your tokens read from cache, with reasoning and cached tokens counted.
When it helps
Choosing a cheaper model
Usage shows how much of your spend goes to each model. You know where a switch saves the most.
A provider having a bad day
The failure rate shows when requests stop getting through. Activity shows which provider served each one.
A bill that doubled
The split by model, key and teammate shows what changed since the week before.
Planning the next top-up
The current pace turns into days of credit left.
Usage questions
How far back does Usage go?
90 days. Older requests drop out of Usage and Activity. Every charge stays on your billing statement, which is never purged.
Which windows can I pick?
24 hours, 7 days, 30 days or 90 days. Empty hours and days show as zero.
Can I see spend per teammate?
Yes. Owners and admins see spend by teammate and by key. Other members see their own keys, their own row and the workspace totals.
Does Usage show latency and cache hits?
Yes. It shows the p95 latency of served requests and the share of tokens read from cache.
Does Usage need my prompts?
No. It's built from request metadata such as model, tokens, cost and timing. By default, Hopscotch never stores your prompts or the model's responses.
Know what every key spent
Give each agent its own key, so Usage can tell them apart.