Last week I decided to actually track my Claude Code token usage in real time instead of just checking the dashboard at the end of the month.
The results changed how I code.
The Setup
I used a macOS menu bar tool called TokenBar that sits in your menu bar and shows live token count + dollar cost as you work. It supports OpenAI, Claude, Gemini, Cursor, OpenRouter, and Copilot.
The key thing: you see spend while you code, not after.
What Surprised Me
1. Context window resets are expensive
Every time I started a new Claude Code session without clearing context properly, the first message was eating 8-12k tokens just on the system prompt + file context. Over a day, that added up to roughly 40% of my total spend.
Fix: I started being deliberate about when to start fresh sessions vs continue existing ones. Saved about $3/day just from this.
2. Sub-agents stack fast
When Claude Code spawns sub-agents for parallel tasks, each one carries its own context. Three sub-agents running simultaneously was sometimes 3x the token burn I expected.
Having the live counter visible meant I could see when spend was spiking in real time and cancel runaway agents before they burned through budget.
3. The "just one more prompt" trap
Without live visibility, it felt like each prompt was cheap. With the counter running, I could see that my "quick follow-up" prompts were often 2-3k tokens each. Ten of those and you have burned $2-4 without building anything new.
4. Model routing matters more than you think
Switching between Opus and Sonnet for different tasks made a massive difference. Opus for architecture decisions, Sonnet for boilerplate. The live cost display made it obvious when I was using the expensive model for simple tasks.
The Numbers
Over 5 days of tracked usage:
- Day 1 (no tracking): ~$18 spent, mostly unaware
- Day 2 (with live counter): ~$14, started noticing patterns
- Day 3: ~$11, actively routing models
- Day 4: ~$9, optimized session management
- Day 5: ~$8, comfortable rhythm
That is a 55% reduction just from seeing the numbers while working.
Why This Matters
Most developers treat LLM costs as a monthly surprise. You open the billing page, wince, and move on. But the waste happens in real time, in small increments that feel invisible.
The fix is simple: make the cost visible while you work.
If you are using Claude Code, Cursor, or any LLM-heavy workflow on macOS, TokenBar is $5 and sits in your menu bar. It changed how I think about token spend.
TL;DR: Tracked my Claude Code tokens for a week with a live menu bar counter. Cut spend by 55% just by seeing the numbers in real time. The biggest waste was context window resets and unnecessary sub-agent spawning.
Top comments (0)