Features
Everything the gateway does
One page per capability, each ending in the concrete behaviour - the status code, the header, the arithmetic - rather than an adjective. Start wherever your question is.
Get an answer, whatever happens
The gateway's core job: pick well, survive a provider having a bad day, and never pay twice for the same context.
The AI gateway
One endpoint in front of seventeen models and seven providers.
Smart routing
Cost, latency, quality or balanced - and the response says why.
Automatic fallback
Cross-provider chains that keep tools and history intact.
Three-layer caching
Exact, semantic and provider caching, with the saving stated.
Know what happened
Every response carries its own receipt, and the dashboard only ever aggregates what your client was already told.
Control the money and the risk
Limits that act before the request runs, and rules that run before any model sees a prompt.
Run it as a team
Who may do what, who approves, and what the record shows afterwards.
Build against it
The parts you reach for once the gateway is already in place.
All of it on one endpoint
Nothing here is a separate product or a separate bill. Point an OpenAI-compatible client at the base URL and every feature above applies to the traffic you already send.