Agents API handles an agent's session and execution loop. Your app still owns admission, bus routing, external authority, self-hosted lifecycle and outcome validation. A completed turn can contain failed tools; cancellation does not establish rollback of external effects.
Applicability and guarantees
Agents API fits applications that want OpenAI to operate the Codex model/tool loop and retain sessions. Configuring it does not replace that harness with the application's OMP or Lisp runtime.
The application still owns who may invoke the agent, which resources a tool may access, communication routing, side-effect reconciliation, and lifecycle of self-hosted compute.
A terminal turn outcome is observable, but completed does not prove every tool succeeded. Cancellation stops the active agent turn; do not infer rollback of operations already performed. Disconnecting an event stream is not cancellation.
A message during an active turn steers it. Changing session instructions or tools requires a new session through the documented update contract.
Private-state limitations and subagent isolation remain as described in the fleet comparison and delegation boundary.