Analytics Dashboard
Tokens, costs, latency. All in one dashboard. Finally, a chart that matters.
Questions this answers
- LLM API cost monitor dashboard
- Visual dashboard for AI agent token usage
- How to view AI coding costs over time?
- Analytics for terminal AI agent API usage
How it works
The Analytics Dashboard aggregates data from token counting, cost tracking, and latency tracking into a visual interface. It presents charts and summaries showing token usage trends, cost breakdowns by provider and model, and latency distributions over time. Data can be filtered by date range, provider, model, agent type, or repository.
The dashboard is accessible through the MCP server, which exposes the underlying data as queryable endpoints. This means you can build custom visualizations or integrate the data into existing monitoring tools if the built-in dashboard does not cover your use case.
Key metrics are highlighted at a glance: total spend over a period, most expensive sessions, token usage growth trends, and latency percentiles. The goal is to answer the question of where your AI budget goes without requiring you to dig through individual calls.
Why it matters
Individual per-call metrics are valuable for debugging, but understanding AI spending patterns requires an aggregate view. The Analytics Dashboard shows token usage, costs, and latency trends over time with daily/weekly/monthly grouping. It queries the same telemetry data that the MCP server exposes, so what you see in the dashboard matches what your automations report.
Frequently asked questions
Can I export dashboard data?
The underlying data is accessible through the MCP server API, so you can query and export it in JSON format for use in spreadsheets, external dashboards, or cost management tools.
Does the dashboard update in real time?
The dashboard reflects data from completed API calls. New data points appear as calls complete, so the view is near real-time during active sessions.
Can I share the dashboard with my team?
The MCP server endpoints that power the dashboard can be queried by any tool or script with access. Sharing the data with teammates or feeding it into a shared monitoring system is straightforward.