See where your
tokens go
You're spending hundreds on structured generation and you have no idea which fields cost what. tokstat reads your JSON corpus, tokenizes every field, and shows you exactly where the money goes.
Per-field token economics
Cost per field
See exactly how many tokens each field costs — avg, p50, p95 across your entire corpus. Estimated dollar cost per model.
Fill rate analysis
Find fields that are null 60%+ of the time. You're paying for the field name and null literal every time — pure waste.
Schema overhead
Field names, braces, brackets, colons, commas — structural tokens you pay for every instance. tokstat shows exactly how much.
Array repetition tax
Arrays of objects repeat every field name per item. 5 items with 8 fields = 40 field names. tokstat quantifies the cost.
Multi-model pricing
Switch between models instantly. Live pricing from OpenRouter. See how costs change from GPT-4o to Claude to Gemini.
Schema cohorts
Mixed-schema corpus? tokstat auto-detects schema variants and lets you analyze them separately or combined.
Four ways to see your data
Area shows relative cost. Click any field to drill into its children. Switch views to spot different patterns.
Automatic insights
tokstat scans your schema and flags the highest-impact optimization opportunities — no manual analysis needed.
Try it now
Two implementations, built in parallel. Pick one and explore.