🛠️ AgentAcct: Observability and Cost Tracking for Coding Agents

As coding agents become a standard part of the developer workflow, a new problem arises: tracking exactly what these AI assistants are doing and how much they cost in API usage. This project addresses that gap by providing a local-first dashboard for tools like Claude Code and Codex. It breaks down each task into granular steps, showing the tools used, files modified, tests executed, and the exact time and tokens spent. With no telemetry and no login required, it focuses purely on local observability and cost tracking.