@DanKornas: Coding agents can say “done.” This repo gives you a receipt. agentacct is a local-first work-intelligence dashboard for…

Coding agents can say “done.” This repo gives you a receipt. agentacct is a local-first work-intelligence dashboard for builders using coding agents such as Claude Code, Codex, OpenCode, and Hermes. It helps you review what happened in each task by turning local session logs and recorded work into a Work Receipt with actions, costs, and evidence kept separate. Key features: • Per-task Work Receipts – capture commands, files touched, tools used, costs, and evidence for a task • Multi-agent coverage – reads local session data from Claude Code, Codex, OpenCode, and Hermes • Evidence tiers – separates an agent’s claim from self-reported checks, observed exit codes, and CI evidence • Local-first design – keeps state in local files, with no account, telemetry, or cloud sync • Multiple ways to inspect work – available in a macOS app, `agentacct tui`, and a local JSON API It’s open-source (MIT license). Link in the reply 👇