agentacct
See what your coding agents did and what it cost. Breaks each task down into work steps — tools used, files changed, tests run, time and tokens spent. Local-fir
What is agentacct?
See what your coding agents did and what it cost. Breaks each task down into work steps — tools used, files changed, tests run, time and tokens spent. Local-first dashboard for Claude Code, Codex, OpenCode, and more. No login, no telemetry.
Overview of agentacct
What it is
agentacct is a local-first dashboard that reads session logs from coding agents (Claude Code, Codex, OpenCode, Hermes, and others) and displays what each agent did and what it cost. It breaks each task into work steps: tools used, files changed, tests run, time and tokens spent. No account, no cloud, no telemetry; all data stays on the local machine.
Key features
- Local-first: reads only local session files, listens only on 127.0.0.1, never stores or requests a provider API key.
- Work Receipts: each task shows a timeline from attempt to result, with recorded steps, checks, and estimated cost.
- Honest labeling: “Reported” (agent’s claim) vs “Verified” (passing checks after last change); estimates marked with ≈; confidence labels on joins.
- Dashboard shows latest recorded work across projects, with client, activity time, outcome, evidence, and cost.
- Filter by date, agent, model, or provider independently; chart shows fresh vs all tokens; tables show fresh, cache, and total tokens.
- Terminal UI (agentacct tui) provides receipts, usage, capacity, and a cross-agent timeline with keyboard navigation.
- Records work (tool names, touched files, credential-scrubbed commands, exit codes, tokens, recorded steps) but never prompts, responses, or transcripts.
Use cases
- Monitor what coding agents are working on across projects and clients.
- Track token usage and estimated costs per agent, model, or provider.
- Verify agent claims by comparing reported outcomes with recorded checks and evidence.
- Compare different agents (e.g., Claude Code vs Codex) on the same task.
- Audit agent activity for billing, accountability, or debugging.
Getting started
bash pipx install agentacct agentacct onboard # once per machine: finds your agents, sets up a local store, starts the recorder agentacct tui # the terminal app Requires Python >= 3.11 on macOS or Linux; Windows supported only via WSL. A signed macOS app is also available from the latest release.
License and pricing
MIT license. Open source.
Based on the project's public README and repository metadata.