What would make you switch to a lower-cost AI API gateway?

I’m launching Frugal Relay tomorrow and would value feedback from developers and technical creators.

The problem I’m trying to solve is simple: many teams want to use GPT and Claude models, but official API pricing, provider-specific integrations, and separate billing setups make experimentation and production costs harder to manage.

Frugal Relay takes the OpenAI-compatible gateway approach: one API for GPT and Claude models, with minimal client-side configuration changes.

Before launch, I’d love to understand:

• What would make you consider switching an AI API gateway?

• Which clients or tools need the best compatibility: OpenAI SDKs, Cursor, Codex, Claude Code, LangChain, or something else?

• Is transparent pay-as-you-go pricing more important than model breadth, reliability controls, or usage analytics?

• For creators who publish developer or AI-tool content, what makes a referral program genuinely worth sharing with an audience?

I’ll use the feedback here to prioritise what we improve after launch. Thanks in advance.