A usage tracking tool for AI agents, splitrail, has been updated to fix two critical accounting bugs. The first bug involved Claude Code's session files being rewritten in place, causing historical usage data to be lost and leading to inaccurate totals. This was resolved by implementing a persistent SQLite history store. The second bug stemmed from splitrail's file discovery mechanism, which had a depth limit that made it invisible to subagent transcripts, resulting in significant undercounting of messages and associated costs, particularly for users who delegate tasks to cheaper models like Sonnet and Haiku. AI
IMPACT Ensures accurate cost tracking for AI agent workflows, particularly for users leveraging Claude Code and delegating tasks to models like Sonnet and Haiku.
RANK_REASON The item describes bug fixes and improvements to a specific software tool (splitrail) used for tracking AI agent usage, rather than a new model release or significant industry event.
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
