
The five views
Overview
Account-level KPIs: total spend, request volume, token consumption, and error rate for the selected period. The starting point for a quick health check.
Cost
Spend over time, broken down by any dimension. Group by model, provider, environment, or any metadata key (agent, customer, feature…) to see where money is going.
Usage
Input, output, and cached token counts. Same group-by options as Cost. Useful for spotting which agents or customers drive token volume.
Performance
Latency and time to first token. Filter to a specific model or metadata tag to compare performance across them.
Reliability
Error rates, status codes, and error types. Shows where requests are failing and which provider or agent is involved.
Group by any dimension
Every view has a group-by control. Pick any built-in dimension or any metadata key you’ve sent:modelorproviderare available on every request.environmentandsession_idappear once you start sendingX-Toolken-Metadata-EnvironmentandX-Toolken-Metadata-Session-Id.- Any other metadata key (
agent,customer_id,feature,tier, …) appears as soon as you send it viaX-Toolken-Metadata-*orX-Toolken-Metadata.
Period selector
Each view has a period selector. Choose Last 7 days, Last 30 days, or Last 90 days. The period applies to the chart, the summary numbers, and any group-by in that view.Reliability: what you see
The Reliability view shows:
Filter by any metadata key (agent, customer, feature) to see whether a specific part of your system is driving failures.
Nothing to configure
The dashboard works from traffic alone. If a request carries no tags, it shows up under model and provider. Add anyX-Toolken-Metadata-* header and that metadata key appears in the group-by controls automatically.
Next
Custom Properties
Add headers to tag requests and unlock every dimension.
Dimensions & Filtering
How grouping and filtering work across views.