Show HN: Agentic Metric – top for your AI coding agents (token, cost tracking)
github.com I use AI coding agents for almost all my daily development. Burning ~100M tokens a day, I had no clue where they all went —
so I built this tool to find out.
Agentic Metric is like top, but for your coding agents. It tracks token usage and costs across Claude Code, Codex, OpenCode,
Qwen Code, and VS Code Copilot, with a live TUI dashboard and CLI.
A few things I care about and you probably do too:
- Fully open source (MIT): https://github.com/MrQianjinsi/agentic_metric
- 100% offline — no network requests, no telemetry, nothing leaves your machine, ever
- All data is local — just reads your local agent files, stores stats in a local SQLite DB
- Read-only — never touches your agent configs or data
What you get:
- Live session monitoring with 1s refresh
- Per-session token breakdown (input/output/cache) + cost estimation
- Today's summary and 30-day trends
- Plugin architecture — easy to add new agents
- Status bar integration (tmux, i3blocks, waybar, vim)
Install: pip install agentic-metric or uvx agentic-metric
I built this entirely with agentic coding (dogfooding at its finest). Would love feedback from fellow heavy agent users —
curious how your daily token consumption compares.No comments yet.