Overview

Token‑Monitor is a desktop widget that tracks token counts, cost, limits and session activity for more than 46 AI coding tools. It works locally on your computer and can sync data across devices through a Cloudflare Worker.

When you run an LLM request, whether you write Python code that calls OpenAI's Chat Completions API, a Node.js script that uses Claude, or a Docker‑based workflow that sends a prompt to Gemini, the widget records the number of tokens received and sent, the monetary cost if the provider calculates it, and any limits on request size or rate. The records are displayed in a table that updates in real time. Beneath the table a trend graph shows total tokens and cost over a configurable period, letting you spot spikes that might indicate a costly query or an error in your code. If you exceed a provider's limit, a warning appears beside the relevant row.

Installation is straightforward: download the pre‑compiled Electron bundle for Windows, macOS or Linux from the GitHub releases page. The app contains a Node.js backend that listens for API calls from the AI tools you use. If you enable sync, the backend emits Server‑Sent Events (SSE) to a Cloudflare Worker; you must create a Worker in your Cloudflare account and provide the endpoint URL in the app's settings. The worker forwards the SSE stream to any other device running Token‑Monitor with the same sync key, so you see the same data on a tablet, laptop or desktop.

The tool is built for developers who frequently run code through multiple LLMs and want a consolidated view of usage. It does not provide alerts beyond the visual warning, nor does it take any action to reduce cost. It also does not bundle any AI models; you must have separate API keys for OpenAI, Claude, Gemini, DeepSeek, etc. And supply those keys in the app's configuration. Because the software is open source, you can fork the code, modify the backend, or run it inside Docker if you prefer containerised environments. The repository includes a Dockerfile that builds a minimal image containing the Electron app and its Node.js runtime.

The listing does not disclose any pricing plans, and there is no mention of user limits beyond the local installation.