Hey everyone,
I’ve been using Cursor with a custom API relay to manage costs across different providers. But I started getting suspicious that my relay was occasionally serving me smaller models during peak hours while still charging the premium rate.
So I built Relaylens, a simple probe tool that tests any OpenAI/Anthropic-compatible endpoint for model authenticity. It works by checking the server-signed thinking signature (a signal an intermediate relay can’t forge).
Why it’s useful for Cursor users:
-If you’re using a custom endpoint in your cursor.json or settings, you can now verify it’s delivering what you pay for.
-Shows you the exact token cost upfront.
I’d love for some folks here to point it at their relays and let me know if it catches anything unexpected.
Also open to feature requests if this is useful to the community 