Local token-cost trackers become a crowded category overnight.

A small tool called Tokentab showed up on GitHub this week and picked up 222 stars in a matter of days. It reads the session logs from Claude Code, Codex, and Gemini CLI right on your machine, no API key needed, nothing sent anywhere else, and it tells you what you’re spending by model and by project. You can check it out at the Tokentab repo.

Here’s the part worth noticing. Tokentab didn’t launch into empty space. At least three other tools, codeburn, TokenTracker, and toktrack, showed up independently around the same time, all trying to answer the exact same question: how much am I actually burning on AI agents right now. Nobody coordinated this. Different developers, same itch, same week.

That’s a signal, not a coincidence. If you’re running coding agents or letting your team loose on Claude Code or Gemini CLI, you’ve probably felt the same gap these tools are filling. You know your monthly bill. You don’t know which project ate it, which model is expensive for what, or whether one engineer’s agent habit is quietly costing more than the rest of the team combined. Right now most owners find out when the invoice lands, not before.

For a small business, that visibility matters more than it sounds like it should. Agent usage scales fast once a team gets comfortable with it, and costs can climb before anyone’s watching closely. Tools like this give you the breakdown you need to catch that early, without adding another vendor or another login to manage.

It’s a small category, but it’s crowding up overnight for a reason. This is the kind of thing we build into an AI command centre, so spend, usage, and project-level cost sit in one place instead of scattered across five different local tools you have to remember to check.