agentacct

agentacct is a local dashboard for developers who want to see what coding agents did across projects and compare their work with resource use. It brings recorded activity and results into task receipts rather than relying on an agent's claim that a job is finished.

• Task receipts: Review participating sessions, tool calls, touched files, recorded outcomes and checks in a task's activity timeline.
• Result verification: Distinguish an agent's reported completion from a verified result, which requires passing checks after the last recorded change; earlier failed attempts remain visible.
• Usage and cost: Filter by date, agent, model or provider, then compare token use and pricing-table cost estimates with the recorded work.
• Project timelines: See sessions from different agents together for a project, with partial coverage and overlapping work labeled.
• Evidence gaps: Confidence labels show how reliably usage connects to work, and missing evidence appears as a gap rather than a zero.

Claude Code and Codex have the fullest task receipts. Support for OpenCode, Kimi Code and Hermes varies by capability. Cost estimates don't use invoices.

The MIT-licensed open source tool runs on macOS and Linux, with Windows support through WSL. It has a terminal interface and a native app for macOS 14 or later. It needs no account or cloud service, has no telemetry, and keeps data on your computer. It records work metadata, including credential-scrubbed commands, rather than conversations.