Why We Added trackmy.codes to Our Stack — Finally Visible Proof of What AI Coding Actually Costs

trackmy.codes ($29/yr) automatically distinguishes 'engine running' from actual coding time across Claude Code, Cursor, and Codex. We found 40% of 'AI hours' were idle. The data changed how we budget.

Published 2026-06-11

Why We Added trackmy.codes to Our Stack — Finally Visible Proof of What AI Coding Actually Costs

TL;DR: trackmy.codes ($29/yr) passively tracks active vs idle time across Claude Code, Cursor, and Codex. We discovered 40% of logged ‘AI hours’ were idle — waiting for reviews, context switches, coffee. The data cut wasteful credit pool burn. Full review →

The Context

Two-dev team on Claude Code credit pool ($100/mo, ~100 credits/day). Pool economics demand discipline: every idle minute burns credits. We had no visibility into actual vs perceived usage. trackmy.codes launched June 2026; installed it to audit.

What We Tested

MetricBefore trackmy.codesAfter (Week 1)Change
Logged ‘AI hours’ (self-reported)~120 hrs/moBaseline
Actual active coding (trackmy.codes)~72 hrs/mo40% idle
Credit pool utilization~95% (frequent alerts)~68%27 pp drop
Opus flag usageAd-hocTracked per-sessionAuditable

The Pivot Point

Day 3: trackmy.codes dashboard showed Tuesday 9am–11am: 120 min ‘session’, 47 min active. The other 73 min: PR review wait, Slack, coffee, context switch to Codex for refactor. That single idle session burned 15 credits ($0.75). Extrapolated: ~$45/mo wasted. We added a pre-session habit: cx/cc launch only when ready to type; close immediately on blocker.

What We Use Now

trackmy.codes (background menubar app, $29/yr):

  • Auto-detects Claude Code, Cursor, Codex, Windsurf processes
  • Distinguishes ‘process running’ vs ‘keystrokes + terminal output’ (active)
  • Exports CSV for monthly audit: date, tool, active_min, idle_min, session_count
  • Team view: shared Notion database with monthly rollup

New protocol:

  1. Check credits remaining (cc usage) before starting
  2. Launch tool → trackmy.codes starts session
  3. Blocked? Close tool. Switch mode if >15 min wait.
  4. Friday 4pm: export CSV → Notion → review idle patterns

When You’d Choose Differently

  • Solo dev, no credit cap: Idle time doesn’t cost cash; tracking is optional curiosity
  • Enterprise with centrally managed IDEs: trackmy.codes is user-space; may not install on locked machines
  • Non-terminal tools (Copilot, Copilot Chat): Not yet supported (roadmap says Q3 2026)

Tool Crucible Rating

DimensionRating (1–5)Notes
Overall4Unique category; fills measurement gap no vendor provides
Ease of Use5Install → forget → review weekly; zero config
Value5$29/yr saved ~$540/yr in wasted credits (our math)
Support3Solo founder; email only; fast but no SLA

This is part of our AI Coding Tool Evaluation series. See full review: trackmy.codes: The AI Coding Time Tracker That Paid for Itself in Day 3

Last reviewed 2026-06-11. See our methodology and affiliate policy.