Three views: Spend, Save, Proof
The dashboard is three views, in the order the questions arrive: where is the money going, what should we change, and did the change work. Each loads in under two seconds.
Spend - where the budget goes
Spend breaks your cost down by model, team and session across five tabs:
- •Overview - total cost and total tokens for the selected period, a daily cost chart, your budget burndown, a "Why Your Bill Changed" explanation and Skill Trends
- •Cost Drivers - the current period against the previous one, with the change attributed to the factors that produced it
- •Models - calls, cost, average tokens and share of spend per model
- •Tokens - input, output, cache-read and cache-write token counts, so cache behaviour is visible separately from fresh input
- •Activity - daily active sessions over the period
Save - what to change next
Save is the work list, in three tabs:
- •Recommendations - advisor suggestions you can accept, with the spend trend they apply to
- •Caching - cacheable patterns detected in your usage data, and what prompt caching would save
- •Prompts - prompt optimization tips that cut token count while preserving intent
Nothing here changes a call on its own. Every item is a suggestion; you decide.
Proof - every saved dollar, traced
Proof shows your verified savings and the math behind them: what you *would* have spent versus what you actually spent after acting on a recommendation, month by month, with the 90/10 gainshare split spelled out. Only recommendations you accepted count - ignored ones claim zero. This is the same calculation Tokeven bills from, which is why it is open book.
The pages around the three views
Separate from the three views, and visited on purpose rather than daily:
- •Team - your organization's members and pending invites
- •Weekly Analysis - the Monday digest built from your own token patterns (Team plan)
- •Board Briefs - a one-page monthly summary of spend, what moved it and the forecast; open a month and print it to PDF (Team plan)
- •Export - period summary plus CSV and JSON export for your own BI tools (Team plan)
- •API & Widgets - the REST API and embeddable widgets for putting Tokeven numbers in your own dashboards
/dashboard itself is the setup wizard - token, install, first tracked call - and stays available after you finish it.
Date range
The range picker lives in the dashboard header and applies across Spend and Save: 24h, 7d, 30d (the default) or 90d. Proof is monthly and does not follow the picker.
What the dashboard can and cannot see
- •Only instrumented traffic appears. Calls made through the SDK wrapper or the local sidecar are tracked. Calls from the Workbench, a console, or an app you have not instrumented are not - which is why a total here can be lower than a provider invoice.
- •MCP alone records nothing. The MCP server is the pre-send advisor: it prices a call before you send it. Ingest events come from
wrap()or the sidecar. Runtokeven statusto see which capture paths are live on your machine. - •No prompt content reaches Tokeven on the ingest path. Every number above is derived from token counts, model IDs and timings.