Token Log
The Token Log is the second tab on the Monitor page. It shows how much the Token Optimizer is saving across your servers.
Hero cards
Section titled “Hero cards”Four cards at the top summarize your savings:
- Tokens Routed: total tokens that passed through the optimizer pipeline
- Tokens Saved: tokens avoided by the optimizer (accent-colored)
- Savings Rate: a ring gauge showing the percentage of tokens saved (0-100%)
- Cost Avoided: estimated money saved, calculated using the model pricing you select
Model selection
Section titled “Model selection”Click the settings icon to choose which AI model to use for cost estimation. Options include Claude Opus, Claude Sonnet, Claude Haiku, GPT-4, GPT-4.1, GPT-4.1 mini, and Gemini 2.5.
The cost calculation multiplies tokens saved by the selected model’s per-token price. This gives you a rough estimate of what the optimization is worth in API costs.
Time-series chart
Section titled “Time-series chart”A line chart shows token savings over time. Toggle between 24-hour, 7-day, and 30-day views to see trends.
Optimizer breakdown
Section titled “Optimizer breakdown”A bar chart shows how much each optimization technique contributed:
- Schema Cache
- Session Dedup
- Lazy Loading
- Null Stripping
- TOON Encoding
- Result Caching
- Budget Truncation
- Description Truncation
This helps you see which optimizers are doing the most work and decide which to enable for other servers.
Per-server savings
Section titled “Per-server savings”A table shows the top 10 servers by tokens saved, with sparkline charts showing the savings trend for each one.
Data freshness
Section titled “Data freshness”The Token Log updates in real time. As tool calls flow through the optimizer pipeline, the hero cards and charts reflect the latest numbers without refreshing.
Usage data is stored in hourly buckets, rolled up to daily (7 days) and monthly (90 days). Older data is aggregated to save space while keeping long-term trends visible.