Model Waygate
One API for every model provider, on your own keys.
Add the credentials you already hold with OpenAI, Anthropic, Gemini or any OpenAI-compatible server, mint a gateway key, and point any OpenAI or Anthropic SDK at the gateway. Each request follows the route you draw for its endpoint to a provider, under your own account with them, in whichever dialect your SDK speaks: the gateway translates between them, keeps the usage and holds the caps.
What it does
Your own keys
Provider credentials are stored encrypted and used only to forward your requests. Prompts and answers are never stored; the usage log keeps the model, status, token counts and timing.
Routes you draw
Each endpoint gets its own URL and a route drawn in the dashboard's editor: providers with their models fixed, fallbacks that move to the next provider when one fails, and decisions where a model you pick chooses the branch. Every request records the way it went.
Any dialect, any provider
OpenAI Chat Completions, the Responses API, Anthropic Messages and embeddings, each answered by whichever provider the endpoint's route reaches, streaming included. A field that cannot be carried across is refused by name, never dropped.
Caps and usage
A monthly spend cap per account and, if you want, per key, with a per-minute rate limit. The dashboard shows the month's spend against the cap, a rollup by endpoint and model, and the latest requests with the route each took.
Priced by the byte
$0.01 per MB of request and response carried for an answer a provider gave, metered through Polar and invoiced monthly. Every response says what it cost. A request that fails before its answer starts costs nothing, a stream is billed for what it sent once it has started, and the provider bills you for tokens directly.