KeyoAPI

I kept hitting the same wall: one project needs GPT-class text, another needs Claude-class reasoning, a third needs image or voice output — and every provider wants its own account, billing and SDK quirks. KeyoAPI started as an internal gateway that routes all of them through one OpenAI-compatible endpoint, then grew a transparent pricing page where you can compare per-token rates across model families before spending a cent. Happy to answer anything about the gateway, and feedback on the pricing layout is very welcome.

Nice launch. Keeping the OpenAI SDK and only swapping the base URL makes trying this very low-friction. Curious: how do you handle fallback when one model provider is down?

The 'keep your SDK, change the base URL' pitch is the right DX. One thing I'd want before routing production traffic: if an upstream model family is degraded, does the gateway fail over automatically or return the upstream error as-is? A per-model status page would help.

Does the pricing comparison show cached-input rates separately from standard input and output tokens? A calculator with an editable input/output mix and cache-hit rate would help compare the likely cost of a real workload, rather than just the lowest advertised rate.

The “keep your OpenAI SDK, just change the base URL” setup is exactly what makes switching easy — nice choice to put pricing front and center too. A few things I’d want to know before routing production traffic: Model transparency — when you say “GPT-class” or “Claude-class”, are requests going to the official upstream models, or to equivalents? A clear model list with exact versions would build a lot of trust. Reliability — do you offer automatic fallback if an upstream provider is down, and what’s the typical latency overhead vs calling providers directly? Data handling — do you log or retain prompts/outputs? A short privacy/retention note on the pricing page would help teams evaluate quickly. Pricing — the $0.01 per million tokens starting price is eye-catching; it’d help to show which model that applies to, and whether prepaid credits expire. A public status page and per-model usage breakdown in the dashboard would make this an easy pick. Congrats on the launch!

Comparing API pricing across 300+ models in one unified dashboard is super useful for cost optimization! Do you update token rates automatically when providers adjust their tiers, and do you account for cached input token discounts in the comparison? Excellent developer tool!

Since you're proxying Claude-class and Gemini-class models behind OpenAI-compatible /v1 endpoints, how do you handle features that don't map cleanly, like Anthropic's tool-use blocks or Gemini's safety settings? Showing per-model notes on the pricing page about which parameters are passed through vs. dropped would make the fallback-model use case much easier to trust.

Having one prepaid key for GPT, Claude and Gemini models is handy. Do you add a markup on top of each provider's own per-token price, and what happens if one model goes down mid-request?

Routing everything through one OpenAI-compatible base URL is the right abstraction — swapping providers without touching client code is where most teams lose a day. Before moving production traffic I’d want to know two things: can I see latency and error rates per upstream provider, and on an outage do you fall back to a model I pick or return a 5xx to my app? A small per-model status page would build a lot of trust for a prepaid gateway.