StackMap
Subscribe

agentacct vs tare

Local-first work receipts for coding agents: reads the session logs Claude Code, Codex, OpenCode and Hermes already write and reports what each task did, what it cost, and whether a check proved it. — versus — Claude Code skill that reads your local session logs and answers 'where did my tokens go' in plain English: deduplicated totals, cost charged to the tool that caused it, 5-hour window state.

The curated verdict

Both turn session logs into receipts without sending anything off the machine. agentacct reports what a task did and whether a check proved it across harnesses; tare answers the narrower question of why your limit hit.

agentaccttare
Stars681161
Forks758
LanguagePythonPython
LicenseMITMIT
Last activitytoday3 days ago
Topicscoding, localcoding, skills, local
Curated connections87

agentacct — the curator's take

The only tool in this neighbourhood that refuses to take the agent's word for it. Everything else here counts tokens and dollars; agentacct counts work and evidence, and keeps them on separate axes — an agent saying 'done' files the task under Reported, and only a machine check that postdates the newest recorded work earns Verified. Every figure wears its basis (`≈` for estimated, bare ` only for reported), every field carries provenance (client hook, transcript scan, MCP record), and unconnected verifier sources stay honestly labelled 'not connected' instead of quietly scoring zero. macOS app, `agentacct tui`, and a loopback-only JSON API; no account, no telemetry, never touches your provider key. Use it when you are accountable for agent output. Overkill if you only want to know this month's spend.

tare — the curator's take

Install it the day you get rate-limited and can't explain why. Two things make it better than the usual token dashboard: it deduplicates Claude Code's repeated response records (~86% inflation on the author's data), and it charges tokens to the tool that *caused* the re-sends rather than the one that returned bytes — which is how one big early file read quietly dominates a week. It is diagnosis, not control: no budgets, no blocking, no compaction, nothing changes your behaviour for you. Claude Code only, so multi-harness shops want a cross-CLI ledger instead. Everything stays local, and SECURITY.md tells you which greps verify that.