How AI billing works:
Every message you send costs input tokens (your prompt + all context) and generates output tokens (the AI's response).
Input tokens include the entire conversation history, all files read, and tool results — this is why long sessions get expensive fast.
Output tokens (code, explanations) cost 3-5x more than input. Caching lets the AI reuse previous context at 90% discount instead of re-processing it.
Cost by Operation Type
hover for info
Operation Details
Tool Usage Frequency
Tool API Equivalent Attribution
Avg Tokens per Tool Call
Redundant Patterns
Model Recommendations
per project
Provider Comparison
multi-provider
--
Optimization Score
Total Potential Savings
--
Plan Optimization Advisor
per provider
Avg Cache Hit Rate
--
Median Session
--
tokens
P90 Session
--
tokens
P99 Session
--
tokens
Avg Messages/Session
--
Session Depth Breakdown by message count
Token Accumulation Curve avg per message index
API Equivalent per Session Depth
Session Survival Curve sessions still active at msg N