fleet-auditor

Token waste across multiple agent platforms is audited by detecting installed systems, collecting usage data with adapters, and storing runs in a SQLite fleet.db to calculate costs and surface waste patterns.

Token cost savings are identified by computing costs using per-token pricing and analyzing historical usage data to detect idle tokens, allowing you to optimize configurations and reduce spend across your agent fleet.

Scanning token usage over a specific lookback window is fully supported, allowing you to configure the time frame for data collection and audit historical runs to analyze waste patterns accurately.

The cross-system token audit supports Hermes and OpenCode platforms, alongside Claude Code, OpenClaw, NanoClaw, and Ironclaw, by using dedicated adapters to collect usage data from each environment.

Visualizing token waste findings across an agent fleet is best handled by the built-in dashboard, which presents cross-system findings and actionable cost-saving recommendations after the audit phase.

To detect installed agent systems before running a token waste audit, you need the agent platforms actively running in your environment so the detection phase can identify them and apply the correct data collection adapters.