StackMap
Subscribe

agentacct vs claude-code-karma

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 — Local-first dashboard over ~/.claude — sessions, timelines, per-session costs, tool/agent/skill/plugin analytics, live activity via hooks, and ticket linking. No cloud, no telemetry.

The curated verdict

Same local-first, no-cloud posture over ~/.claude with timelines, per-session costs and tool/skill/plugin analytics. Karma is Claude Code-centric telemetry; agentacct spans Codex, OpenCode and Hermes too and grades evidence rather than activity.

agentacctclaude-code-karma
Stars667320
Forks7532
LanguagePythonPython
LicenseMITApache-2.0
Last activitytoday3 days ago
Topicscoding, localcoding
Curated connections74

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.

claude-code-karma — the curator's take

Your ~/.claude directory is a goldmine Claude Code never shows you — Karma turns the JSONL into browsable sessions with full chronological timelines, cost and cache-hit breakdowns, subagent trees, and a surprisingly complete inventory of every plugin, skill, and hook you've installed with usage stats. Read-only ticket linking (Linear/Jira/GitHub) is the team-ready touch. Know the limits: Claude Code prunes session data after ~30 days, so history evaporates unless you keep Karma indexing; it's a two-process dev setup (FastAPI + SvelteKit), not a single binary; and it's Claude-only — no Codex/OpenCode. Pair with ai-token-monitor when all you want is the spend number next to your clock.