New post: I counted my tokens before they hatched.
A look at tokscale, the CLI that tells you exactly how much your Cursor + Claude habit costs.
https://t.co/p7XuCfEgVW
h/t @_junhoyeo for making it
Tokscale is now on X.
What began as a simple open-source usage tracker now tracks trillions of tokens for developers around the world.
Follow @tokscale for product updates, usage insights, and what we’re building next.
Tokscale has passed 5,000 stars on GitHub.
Track token usage and cost across AI coding agents, with public leaderboards for developers and teams.
https://t.co/qBuEA9rsq1
Tokscale now supports @PrimeIntellect’s Prime Agent.
Track token usage and costs across root sessions, forks, and RLM subagents using only your local session data.
It understands Prime’s append-only session trees:
• Root sessions
• Forked history
• RLM child sessions
• Per-model token and cost accounting
• Workspace and agent attribution
Fork and child usage is reconciled to prevent double counting.
npx tokscale@latest
tokscale just got a lot better 🛰️
> Full @opencode 1.2+ support: SQLite-first, zero duplicate counts
> `--today` actually means today (timezone-aware, finally)
> @OpenAI Codex archived sessions counted
> @cursor_ai model pricing as a first-class source:
→ GPT-5.3 Codex now resolves correctly at $1.75 in / $14 out
> Native binaries now live in 8 separate platform packages
> Supports Kimi Code CLI @Kimi_Moonshot
Big week for tokscale
+18k lines added, -1k removed
→ @badlogicgames Pi agent support (badlogic/pi-mono)
→ @OpenClaw session detection & tracking
→ SQLite support for @OpenCode 1.2+
→ 4 new contributors this cycle 🛰️
Elon said "Electricity is money"
Kardashev scale = civilizations ranked by energy consumption
AI turns electricity into value, coding agents turn tokens into code
So I built Tokscale (Token + Kardashev Scale)
I mass burned 7.3B tokens this year.
1B of them belongs to @grok which handles codebase exploration & pattern matching like an SCV scouting the map, blazing fast while completely free.