Volume, cost, latency, and error patterns grouped by model, provider, and owner.
Usage
Requests
Tokens
Input + output
Views
Team + model
New capabilities
Group requests by team and key name to find the handful of workloads behind most of your token spend and request count.
Model
gpt-5.5
Request volume and spend over the selected range.
Provider
Anthropic
Usage grouped by provider route.
Team
Support
Owner-level view for reviews.
See which model costs more for the same work, so a model choice is backed by your own numbers instead of a vendor benchmark.
Spot rising volume, slower responses, climbing cost, or a creeping error rate over time, while there's still room to change a route or fix a prompt.
Who Concentrate is designed for
Request logs are one row per call. Usage analytics roll those rows into comparisons and trends — which models drive the bill, which keys are growing — so engineering and finance can decide what to change without exporting spreadsheets by hand.
Compare models on cost and token volume before changing a route or prompt.
See which teams and keys drive cost over a billing period without opening separate provider consoles.
Spot a creeping error rate or traffic shift early, while you can still reroute or cap a key.
Every chart traces back to request logs, so totals always reconcile to real calls.
Feature basics