Cua's model router puts provider selection and failover behind one service credential. At its November 2025 launch, the service listedAnthropic's Claude Sonnet 4.5 and Haiku 4.5 as its launch models, routing requests among their hosting providers. Developers selected the model while Cua handled the upstream connection.
The same Cua key could access both model inferenceRunning a trained model to get answers — the phase where AI is actually used, as opposed to trained.Full definition → and desktop sandboxes, reducing the separate accounts an AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → application had to manage. Responses itemized Cua's gateway cost and the upstream model cost, allowing developers to see the two charges within the combined service.
Terms in this piece · Glossary
inference — Running a trained model to get answers — the phase where AI is actually used, as opposed to trained.
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
