Hardware project Clawdmeter tracks Claude Code token usage in real time
Reykjavik-based developer Hermann Haraldsson releases a Bluetooth-connected device to visualise Anthropic’s AI coding tool consumption, sparking interest in the niche practice of maximising AI token utilisation.
Hermann Haraldsson, a software developer based in Reykjavik, Iceland, has released Clawdmeter, an open-source hardware project that displays Claude Code usage statistics on a compact desktop dashboard. Launched on GitHub on May 10, the device utilises a Waveshare ESP32-S3-Touch-AMOLED-2.16 display and connects to laptops via Bluetooth, allowing users to monitor their consumption of Anthropic’s AI coding tool in real time.
The project has garnered over 800 stars and 50 forks, reflecting a broader trend among software engineers known as "tokenmaxxing." This practice involves monitoring and maximising the number of AI tokens consumed during work, serving as a metric for how thoroughly developers have integrated AI into their workflows. Haraldsson describes the device as a "fun side project" rather than a critical productivity tool, designed to appeal to AI power users through its interactive interface.
Clawdmeter features pixel-art animations of the 'Clawd' sprite that intensify with token usage, creating what Haraldsson describes as a "little dopamine loop." The device reads the user's Claude Code OAuth token to make an API call, pulling usage numbers directly from the response headers. A middle button cycles through animations and displays simple charts for session and weekly data, while side buttons send Space and Shift+Tab commands to toggle between Claude Code’s Normal, "Accept Edits", Plan, and Auto modes.
Haraldsson, who is not an embedded developer by trade, used Claude to guide him through the device’s development process over a few days. He noted that AI has "democratised access to programming," enabling individuals without specialised hardware engineering backgrounds to build functional devices. The project’s design focuses heavily on aesthetics, with Haraldsson spending significant time refining fonts, colours, and animations to evoke a sense of nostalgia for dedicated hardware devices like the Walkman or iPod.
The device is powered by a lithium-ion battery and serves as a gamified interface for tracking AI interaction, with one community member likening it to a "hardware Tamagotchi for my context window." While the long-term sustainability of tokenmaxxing as a productivity measure remains unclear, and specific security implications of reading OAuth tokens have not been fully detailed, the project highlights the deepening integration of AI tools within the developer community.
