OpenAI has launched the Agents API, a managed service that lets developers build and deploy cloud-based AI agents. The API runs on the Codex orchestration harness and supports long-running sessions and tool use without requiring developers to manage the underlying infrastructure.
The infrastructure angle is what matters here. Offloading session state and orchestration to a managed layer removes one of the core engineering bottlenecks in production agent deployments. The Codex harness handles the complexity that typically forces teams to build and maintain their own scaffolding.
The full announcement details the specific tool integrations, session limits, and pricing structure that will determine whether this is a serious production path or a prototype sandbox. Those specifics are what you need to read before making architectural decisions.
[READ ORIGINAL →]