What a useful tracker should show
A useful AI coding usage tracker puts activity and capacity in the same view. Tokens and requests explain what happened; remaining quota and reset times explain what you can do next. Looking at only one of those leaves out the planning part.
UsageAtlas combines the supported tools already signed in on your computer into one running total. It keeps the detail behind that total available too:
- tokens, requests, models, and estimated cost;
- provider-reported usage windows, remaining capacity, and reset times;
- daily history, comparisons, and your own 28-day baseline; and
- per-window desktop alerts before a limit becomes a surprise.
Why one dashboard matters
Switching between a terminal status line, an IDE, and separate provider pages makes it hard to see the total cost of a workday. One local dashboard makes cross-tool patterns legible: which tool carried a project, which model produced the load, and when your heaviest hours occur.
UsageAtlas currently brings together Codex, Claude, Cursor, and OpenCode. You can filter to one tool when you need an answer about a specific provider, then zoom back out when you want the whole picture.
Keep usage tracking local-first
Usage data can reveal how and when you work. UsageAtlas is designed to read supported local provider state and session data on your device, then render bounded aggregates in the desktop app. It does not read prompts or responses, and it does not require a UsageAtlas account.
Cost figures are estimates unless a provider reports the cost directly. Use the provider's billing records for purchasing or accounting decisions.
Questions, answered
Which AI coding tools does UsageAtlas support?
UsageAtlas currently supports Codex, Claude, Cursor, and OpenCode. Each source is added only after its sign-in path, error handling, and adapter tests are ready.
Can an AI usage tracker show reset times?
Yes. When a supported provider reports a metered window, UsageAtlas shows the remaining capacity and the local time it resets, and can alert you before the threshold you choose.
Does UsageAtlas read prompts or AI responses?
No. UsageAtlas is built around usage aggregates such as token counts, request counts, models, costs, and provider capacity. Prompt and response content is not read.