Second independent agent cost-tracker goes viral

A small command-line tool called TokenTab is getting attention because it tackles a problem many teams are already feeling: nobody has a clear view of what AI coding agents are costing them.

The tool reads session logs from Claude Code, Codex and Gemini CLI, then breaks spending down by model, project and day. It reached 527 GitHub stars in five days, averaging 104 stars a day, according to the project’s GitHub repository. That is not proof it is the best tool in the category. It is a useful signal that agent cost tracking has become an immediate pain point for developers and operators.

The more important detail is that this is the second independent tool focused on agent spend visibility to gain traction this week. When separate builders chase the same narrow problem at the same time, it usually means the market has moved ahead of the standard reporting. Companies are deploying coding agents, but many still cannot answer basic questions: Which team is spending what? Which models are driving the bill? Which projects are producing enough value to justify continued usage?

You do not need a complex finance process to act on this. Start by making agent usage visible by team, project and model before usage becomes difficult to unwind. Set a regular review point, even if it is only weekly. This is the kind of thing we build into an AI command centre, giving operators one place to see usage, cost and activity rather than chasing logs across separate tools.