MagnumGoYB / aitok

What is aitok?

aitok is a Go-based CLI that tracks how much you're spending on AI coding assistants by reading local tool logs. It pulls usage data from Claude Code, Codex, and Gemini CLI, then spits out token counts and estimated USD costs. No API keys, no cloud sync, no data leaves your machine. You get commands like `aitok summary`, `aitok tui`, and `aitok budget check` to monitor spending across time periods, filter by model or provider, and set spending limits that fail your CI pipeline if exceeded.

Why is it gaining traction?

The offline-first angle is the hook. Developers are increasingly nervous about sending usage data to third parties, and aitok sidesteps that entirely by parsing local JSONL logs. The TUI dashboard gives you a quick visual overview without opening a browser. Budget enforcement in CI is practical for teams watching costs. The ability to override default pricing with custom rates for different providers or accounts handles the reality that many teams use multiple API endpoints.

Who should use this?

Individual developers and teams using Claude Code, Codex, or Gemini CLI who want visibility into token consumption without cloud dependencies. Engineering managers tracking AI tool spend across a team will find the budget checking and reporting useful. Anyone running these tools in regulated environments where data locality matters.

Verdict

aitok solves a real problem with a clean, privacy-respecting approach. The offline design and budget enforcement features are genuinely useful. However, the project is early stage with only 26 stars and a 1.0% credibility score. Test coverage appears solid from what the repository shows, but the low community traction means you're adopting something that hasn't been battle-tested by many users. Worth trying if the offline model fits your needs, but factor in that you're an early adopter.