100 million tokens a day. That's how much one developer says he was burning through AI coding agents before he realized he had zero visibility into where those tokens actually went.
The result is Agentic Metric, an open-source tool (MIT license) that works like the Unix top command but for AI coding agents. It monitors token consumption and costs across Claude Code, Codex, OpenCode, Qwen Code, and VS Code Copilot through a live terminal dashboard and CLI.
The tool fills a real gap. Most AI coding agents show you a running total at best, but none break down usage across sessions, projects, or tools in a unified view. If you're running multiple agents across different editors and terminals, your monthly bill is essentially a black box until it arrives.
Agentic Metric runs locally, reads from each agent's log files, and presents a real-time TUI (terminal user interface) showing per-agent token counts, cost estimates, and usage trends. It's lightweight enough to leave running in a side terminal without impacting your workflow.
The project is available on GitHub under the MIT license, meaning you can fork it, modify it, or integrate it into your own tooling. For teams burning through thousands of dollars in API costs monthly, even a rough breakdown of where tokens go can surface obvious waste - like agents re-reading entire codebases on every prompt when a smaller context window would do.