OpenTokenMini

Know what your AI agents cost, and what you get for it.

Your team uses Claude Code, Codex, Gemini CLI or a mix. OpenTokenMini records what every call costs, how long it takes and where it fails, links that to the commits and pull requests it produced, and shows where the spend goes and what to change. It works the same across agents and models, and it's open source: you can read everything it collects.

Agent spend by vendor, split by the work it went to and what came out of it (illustrative numbers)

Install

Ask your own coding agent to do it:

whoever owns the data: "follow the service section of measure/AGENTS.md". It sets up the service and gives you an address and a key.
each developer machine: "follow the workstation section of measure/AGENTS.md" with that address and key.

Prompts and code are not collected. Nothing is collected before you turn it on.

Results

Measured cases from teams using it, and how to add yours: use case reports

Need help?

Deployment, corporate login, or finding where your tokens go: contact us