← Back to Selected Work

AI Tools & Desktop Integration

AI Usage Tracker

Local App · Partial Integrations

A local Windows dashboard for AI usage limits and reset times, with stale readings and incomplete connections clearly marked.

Local App · Case Study

Capability illustration · no account or usage data

Key DecisionKeep a missing or stale reading visibly different from available quota.

The Case Study

Problem
Working across AI tools means checking separate usage limits, reset windows, and dashboards. A missing or stale reading can easily look like available capacity unless the interface makes uncertainty explicit.
What I Built
A Windows sidebar and browser dashboard with separate quota windows, reset countdowns, usage history, connection details, and settings. The desktop launcher starts the local service and restores the existing app when opened again.
How It Works
Provider adapters put readings into a common model while retaining their original timestamps and units. Usage data stays on the device. The Electron app and browser share one interface. Codex readings use its installed account interface; the other provider connections are still being validated. Unknown and expired values remain explicit.
Where It Stands
The Windows and browser app is usable, with Codex readings reconciled against their provider source. Claude and Cursor live reconciliation remain pending. Public source is available; account data stays local.
Evidence and Limits
  • Codex has a verified live reading. Claude and Cursor are incomplete integrations, not confirmed live connections.
  • Provider observations retain their timestamps and units. Subscription prices are user-entered and do not establish the size of a quota.
  • The visual is a capability illustration with no account or usage data. The dashboard does not run an AI model.
Product Stack
React · TypeScript · Electron · Node.js · Zod · Provider Adapters · Browser Extension
Evidence As Of
2026-10-27
Explore All 11 ProjectsGet in Touch ↗