Tiny Agent ToolsEARLY BUILD
← All tools
Logs & Sharing

Usage Log Analyzer

Normalize Claude stream token usage with duplicate and unknown-coverage reporting.

Your input stays in this browser.No input upload. No local history saved.
256 KiB limit

Input

Common subset · 2026-10-06

Result

A useful result starts here.

Paste your input or load an example,
then run the tool.

Supported formats & limitations
  • Processed entirely in a browser worker; no input upload or local history.
  • Rules snapshot: 2026-10-06. Inputs are bounded to 256 KiB and structural/record limits apply.
  • Latest result modelUsage per session replaces assistant counts; never sum cumulative results. Use one cumulative stream per session. Resets must have distinct session IDs; older exports that restart counters need separate analysis.
  • Fallback: main-loop assistant IDs deduplicated using per-component maxima. Assistant output counts are placeholders and remain unknown. Partial events and subagent assistant records are omitted.
  • Missing or unsafe counts stay unknown. Unknown fields are outside the counting subset. No invoice accuracy claim.
  • Export normalized-usage.json for the API Cost Reconstructor. Cache counts exclude ordinary input tokens.
  • Source: https://code.claude.com/docs/en/agent-sdk/cost-tracking

Execution budget: 5s; output limit: 2048 KiB. Browser worker startup has a separate 3s allowance.