Why
Usage Radar is more useful when it can cover the AI tools people actually switch between during the day.
The app already supports Codex, GitHub Copilot, and OpenCode Go. The next provider work should build on that foundation without weakening the tray-first experience or pretending confidence where the source is incomplete.
What I want
Add more providers in a controlled way.
Good candidates can include:
- Claude Code
- Cursor
- OpenAI API or credit balance
- other coding-agent tools with a reliable usage source
Each provider should have a clear source, confidence level, reset timing behavior, and degraded state before it is treated as supported.
Done when
- at least one new provider is added behind the same provider snapshot model
- the provider has an honest confidence label
- stale, partial, and unavailable states behave consistently with existing providers
- the home view and tray state continue to stay compact
- provider-specific auth or setup does not leak secrets into cache or logs
Notes
This should happen provider by provider. I do not want a broad provider list if the data is weak or the UI starts feeling like a dashboard.
Why
Usage Radar is more useful when it can cover the AI tools people actually switch between during the day.
The app already supports Codex, GitHub Copilot, and OpenCode Go. The next provider work should build on that foundation without weakening the tray-first experience or pretending confidence where the source is incomplete.
What I want
Add more providers in a controlled way.
Good candidates can include:
Each provider should have a clear source, confidence level, reset timing behavior, and degraded state before it is treated as supported.
Done when
Notes
This should happen provider by provider. I do not want a broad provider list if the data is weak or the UI starts feeling like a dashboard.