Prompt Inspector: stop overpaying for prompts a cheap model could nail
An OpenAI-compatible proxy that inspects each prompt with a cheap model first, then routes it to the least expensive tier that can actually handle the job.
Prompt Inspector is a small piece of infrastructure with an opinion: you are almost certainly overpaying. Point your OpenAI-compatible client at it, and before a prompt hits an expensive model, a cheap model inspects it and decides which tier can actually handle the request.
Trivial prompt? It goes to a cheap tier. Genuinely hard reasoning? It gets escalated. The cost savings come from the prompts that never needed a frontier model in the first place — which, once you measure it, is most of them.
Why build it as a proxy
Because a proxy is a drop-in. You don't rewrite your app to save money; you change a base URL. Keeping it OpenAI-compatible means the existing SDK, existing code, and existing habits keep working — the routing just happens underneath.