Image: GitHub
Why it mattersA developer paying for several AI tools can see what is left in one place. The vendor dashboards stay authoritative on billing.
A developer paying for Claude, Codex and Cursor sees three dashboards and no single number for what is left. TokenBar reads them all into one tray icon.
TokenBar is an open-source monitor for over 50 AI services. The project is licensed MIT, built on Flutter, and maintained by GitHub user lwh0630. The repository holds 87 stars three days after its first commit on 24 October.
What it watches
The README groups the services into four lists. Three are subscription plans: ChatGPT and Codex, Claude Pro and Max, and Cursor. Twenty-three are official API keys. The vendors covered include OpenAI, DeepSeek, Moonshot, xAI and OpenRouter. Fireworks, Hugging Face, ElevenLabs and Vercel AI Gateway are on the list too. Eight are coding plans such as z.ai GLM and Kimi Code. Sixteen are self-hosted services like LiteLLM and Bifrost. The project reads each one through that service's own API.
What a user sees
The tray icon shows a two-bar indicator picked from the tightest quota window at that moment. The main dashboard lists every account with its percentage left and its reset countdown. The windows it tracks are 5-hour, weekly and monthly, the README says. Usage history draws stacked bar charts over 24 hours, 7 days or 30 days. Low-quota notifications fire at thresholds the user sets.
TokenBar runs on Windows 10 and 11, macOS 12 and above, and Linux (listed as experimental). The mobile side runs on Android 7 and above, and on iOS 15 and above by sideload. The README says credentials stay on the device in that platform's secure storage. An optional sync server moves usage snapshots across machines but never the key.
The maintainer's "fictional demo data" screenshots show a Windows tray panel and a mobile home screen. There are also usage history charts and a desktop dashboard. Six people have forked the project.
Where it fits
A developer who pays for Claude, Cursor and Codex has a known problem. Each plan counts against its own clock. Running out mid-afternoon on a tight deadline is a small daily cost. The vendor dashboards do not solve it, because each dashboard shows one plan. The practical answer is a tool that reads them all and shows the smallest number first.
This is not the first tool in that shape. Pulse covers subscription dashboards on macOS. Codenotch does the same job on Mac's menu bar. TokenBar's claim is width. The list covers 50 services across subscriptions, API keys, coding plans and self-hosted gateways. It runs on five platforms, including a phone. The test is whether each per-service reader keeps working over time. Quota endpoints are not documented at most of these vendors. Each reader can break when the vendor releases a new billing page.
The project does not claim billing accuracy, and does not need to. The vendor's own invoice stays authoritative. What TokenBar offers is a tray icon that goes amber before a developer notices they are about to run out.
Source
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.