Cost Tracking

Tracks token usage and estimated costs across 30+ LLM models. Alerts when costs exceed budgets or usage patterns suggest inefficiency.

Examples

• Agent trace consumed $4.50 in API costs, exceeding the $2.00 budget
• Agent using expensive model for tasks that a cheaper model could handle
• Total token usage for a simple task exceeds 100K tokens
• Cost trending upward across sequential agent steps

Detection methods

Per-Model Pricing

Tracks costs for 30+ models across 8 providers

Budget Comparison

Alerts when trace costs exceed configured thresholds

Model Resolution

Maps model version strings to canonical pricing entries

I/O Separation

Distinguishes input and output token costs