OpenAI brings the Codex harness to a managed Agents API.
OpenAI launched its Agents API in public beta, combining a managed Codex harness with configurable execution environments, context management and subagents. Teams still own the domain workflow and its outcomes.
OpenAI introduced the Agents API in public beta on September 10. It provides managed access to the harness behind Codex, while developers select the task, tools and execution environment. The announcement describes OpenAI-hosted sandboxes, customer infrastructure and partner environments.
The service includes context compaction, tool discovery and parallel subagents. These capabilities address recurring infrastructure problems in long-running agent systems. They are distinct from the OpenAI Agents SDK, a library developers can use to build and control their own agent applications.
Our view: a managed harness can reduce infrastructure work, but the organisation still needs to decide what counts as success and which actions require human review. Pilot a bounded process with representative cases, explicit permissions and recoverable actions. Compare reliability and total cost with your current system before expanding the agent's authority. Public beta also means teams should expect the product to evolve.
buy the runtime when it frees you to improve the actual work.
Read the primary source. Editorial interpretation is our own.