Model per call
You select the model on every request. Router, guardrails, knowledge and Memory stay under the same key and the same organizational policies.
AI Router
One API for every model call
The AI Router serves the full managed model fleet from your regional endpoint: LLM, vision, embedding and speech under one API key, one set of guardrails and one governance control plane.
How routing works
Routing is a control-plane capability: the endpoint stays in your region, and every call passes the same identity, guardrail and knowledge checks as any other platform request.
OpenAI-compatible v1
The regional v1 API speaks the same request and response schema as the OpenAI API, so existing OpenAI clients can point at your endpoint with the same key format.
Region-pinned by design
The endpoint in the URL names the region that runs the call. Routing never moves a request outside the selected market.
Start on the router
Point your existing OpenAI client at your regional endpoint and the managed model fleet is one API key away.
Related platform capabilities
Models & Inference
Region-pinned inference across the live model catalog, under one SLA.
Available now
Custom Models & Training
Train and fine-tune models on your own dedicated capacity.
Coming soon
Agents, MCP & Tools
Build agents with tools, MCP access and multi-step execution.
Coming soon