Use Relixr with Cursor
Point Cursor at Relixr for every model behind one key, with a hard spend cap instead of a rate limit.
Cursor Pro throttles you when you hit its usage limit. Relixr gives Cursor an OpenAI-compatible endpoint instead — every model Relixr supports, billed per request, with a hard cap so a runaway agent loop can't outrun your budget.
This is not an either/or. Keep Cursor Pro for its editor experience — point it at Relixr as the model provider so you're never blocked by its built-in rate limit.
1. Create a scoped key
Create an inference key named after the tool that will use it (cursor), so a usage export six months from now tells you exactly what spent the money. From API Keys → Create key, set a daily spend cap you're comfortable with — this is the thing Cursor's own subscription doesn't give you.
2. Point Cursor at the gateway
In Cursor: Settings → Models → Override OpenAI Base URL. Set:
Cursor will now list whatever models your key allows. Pick one as your default.
If Cursor caches the model list, reopen the model picker or restart Cursor after saving the base URL.
3. Verify it's working
Send one message in Cursor's chat, then check Usage in the Relixr dashboard — the request should show up within a few seconds with the model, cost, and latency.