tokscale

Verdict: worth adopting if you need visibility into AI coding spend, but it should be understood as a usage-tracking and visualization layer for AI coding agents—not a general FinOps suite, not a model gateway, and not an automatic cost optimizer. Popularity is well supported by GitHub stars and social reposts; usability is better supported by the official repo, user screenshots, and practical write-ups discussing Cursor/Claude Code cost tracking.

Its practical value is straightforward: it consolidates token usage and estimated cost from multiple AI coding agents into one terminal/TUI view, so developers can see which tools are consuming the most. Evidence explicitly mentions Cursor, Claude Code, Codex, Gemini, and others, and some users share multi-machine histories. A better analogy is “a local dashboard and ledger for AI coding-agent usage,” not an AI IDE or an agent framework.

On effort and cost, the strongest supported claim is that it can be run via npx or bunx, suggesting a low setup barrier for developers, and the repository is open source. But the money figures it shows appear to be estimated from API pricing rather than guaranteed billing truth.