Token Receipts Audit

•
19
•

Audits LLM usage by both token volume and dollar spend to uncover the true cost drivers across your models and agent workflows.

Token Receipts Audit helps engineering teams analyze and attribute AI spend accurately by separating raw token volume from actual dollar expenditures. Because high-volume open models often dominate agent execution loops while premium models capture the bulk of total costs, simple token-ranking dashboards frequently lead teams to optimize the wrong workloads. This skill tiers models into cheap and premium buckets, calculates exact cost-volume splits from real usage exports, and provides a clear breakdown that reveals whether high-tier spend is justified by complex tasks or wasted on bulk operations.

Key Features

01Model tiering separating economical open models from premium high-cost models
02Defensive guardrails preventing misleading conclusions from volatile short-term windows
03Built-in verification script to validate receipts and compute exact cost percentages
0419 GitHub stars
05Dual-metric breakdown attributing usage across total token volume and dollar costs
06Routing anomaly detection to identify bulk tasks unnecessarily sent to premium models

Use Cases

01Benchmarking LLM routing efficiency to identify candidates for model downgrading
02Auditing multi-agent architectures to distinguish high-volume background loops from expensive reasoning calls
03Investigating unexpected spikes in monthly AI gateway and API provider invoices