A Physical Dashboard for Tracking Claude AI Coding Assistant Usage

Hermann Bjorgvin has built Clawdmeter, a small physical screen that sits on your desk and shows you how much of your Claude Code quota you've used. The device runs on an ESP32-S3, a common microcontroller board, paired with a 2.16-inch touch screen. Instead of checking your usage stats in software or through a web interface, you get a real-time display right in front of you.

What's Under the Hood

The Clawdmeter uses a Waveshare ESP32-S3-Touch-AMOLED-2.16 development board. Think of a development board as a small, programmable computer with built-in networking and the ability to drive a screen. The ESP32-S3 has two processor cores running at up to 240MHz, built-in Wi-Fi, and enough connections to add sensors or controls later if needed.

The key piece here is the screen itself. It's an AMOLED display—the same kind of technology used in high-end phone screens. AMOLED is bright and uses less power when displaying dark backgrounds with bright text, which is ideal for a dashboard that shows numbers and updates throughout the day. The screen is also touch-enabled, so you could add buttons or controls without needing physical buttons.

How It Gets the Data

Clawdmeter reads usage information directly from your computer's hard drive, specifically from the folder where Claude Code stores its data. When you use Claude Code, it automatically tracks how many tokens—small units of text that the AI processes—you've consumed. Clawdmeter watches that data folder and pulls the latest numbers.

This local-file approach has an advantage: once set up, the dashboard doesn't need internet to work. It reads data that's already on your machine. However, getting the data from your computer to the ESP32 device requires sharing files over your home or office network, typically through standard file-sharing methods like SMB or NFS (the same tech that lets you access shared folders on a network).

What You Actually See

The small screen is designed to show your usage metrics at a glance. You might see how many tokens you've used in the current session, your daily total, or trends over time—whatever data Claude Code stores and makes available. Because AMOLED screens are crisp and readable even in bright rooms, the numbers stay clear whether there's sunlight on your desk or just your monitor's glow.

Why a Physical Screen Matters

Most software tells you metrics through notifications or by opening a window. With Clawdmeter, the information is always there, in your peripheral vision. You build awareness of your usage without having to switch windows or interrupt your coding.

There's a pattern here worth noting. When cloud services like AWS became common in the 2010s, developers wanted the same kind of persistent visibility into how much computing power they were using—and what it was costing them. A physical dashboard serves a similar purpose. With AI tools becoming standard in development, tracking token usage (which translates to cost or quota limits) becomes something developers want to keep top-of-mind. A dedicated screen handles that more effectively than software notifications, which compete for attention with all your other tools.

How It Works Technically

The ESP32 likely polls the data folder on your computer on a regular schedule—checking for updates every few seconds or minutes. To make this work, either your computer runs a simple file server that the ESP32 can access, or the ESP32 connects to a shared network folder directly. This introduces standard networking considerations: the connection needs to be reliable, and there may be authentication involved to keep the data secure.

The benefit of the ESP32-S3's dual-core design is that one core can handle network tasks while the other manages the display, so pulling new data won't cause the screen to freeze or become unresponsive.

Bigger Picture

Physical dashboards for tracking digital services are becoming more common as cloud and AI tools hide resource use from view. Clawdmeter is an early example—the code is open-source on GitHub, so developers can modify it to track other AI services or add additional metrics. The same basic approach could work for monitoring API usage, cloud spending, or other consumption-based services.

Over time, this kind of usage tracking will likely move into the integrated development environments and coding tools themselves. Claude Code or similar tools may eventually build this visibility directly in. But dedicated hardware has staying power for one reason: it solves a real workflow problem that software-only solutions struggle with. A screen on your desk is always visible; a software alert has to compete for your attention.

This project also underscores a broader point about hardware and software. Even as software capabilities expand and become more capable, there are still tasks where a small, dedicated, single-purpose piece of hardware does the job better. The ESP32 isn't doing anything fancy—it's just reading a folder and displaying numbers—but for that specific purpose, it works well.