Israeli Developer Builds Analog Gadget to Track Claude Token Limits
Israeli DevOps engineer Orel Fichman has developed CodePanion, a physical device featuring an analog voltmeter that provides real-time tracking of Claude Code token usage.
From the CodePanion repo
Anyone who works with Claude Code knows the frustration of hitting a token limit right in the middle of a critical task. Instead of settling for a boring error message, Israeli DevOps engineer Orel Fichman found a creative way to track his token usage with a custom-built gadget.
An homage to the analog
"It all started when I ran out of tokens in the middle of a task," says Fichman, an engineer at the Israeli startup Dream. "While waiting for a budget increase, I thought about how to avoid this again, but in a way that was truly over-engineered."
Fichman decided to model, 3D print, and build a component that displays remaining tokens on a physical gauge. "There are already 1,001 software-based projects for this, but they were all too elegant. I wanted something a bit more ridiculous," he explains regarding his creation, CodePanion.
As an homage to analog tools, he opted for a needle-based voltmeter. To make the device feel "alive," he added moving eyes to the logo that react to various system states, from idle to tool execution.
A genius-stupid method
Since direct API monitoring was not straightforward, Fichman used a method he defines as "as genius as it is stupid." The architecture relies on a Python daemon running on a Mac that intercepts Claude Code hooks.
"We take the Claude Code OAuth token, send a request to the API, and limit output tokens to one to avoid wasting quota. We are only interested in the response headers, which contain a number between 0 and 1 indicating the percentage of the quota used," Fichman explains. This data is sent to an ESP32 board, which converts it into a 0-3 volt signal for the analog voltmeter.
The device casing was designed in Fusion 360. Fichman notes that this was the most challenging modeling phase, where he relied solely on his own engineering skills rather than AI. The source files for building your own "Claude gauge" are available on GitHub.
