Summary
Alongside the OpenAI Agents SDK launch on September 10, 2026, Temporal introduced agentic sandboxes that pair its durable execution engine with isolated sandbox environments for OpenAI agents. The combination gives long-running agents crash recovery and replayable state while they execute code in isolation.
What changed
Temporal introduced agentic sandboxes integrated with the OpenAI Agents SDK, combining durable, replayable workflow execution with isolated sandbox environments for agent code execution.
Why it matters
Durable execution addresses a core weakness of long-running agents, which is surviving crashes and resuming deterministically, and pairing it with sandboxing targets teams running agents as production workloads. It extends Temporal from microservice orchestration into agent infrastructure.
Evidence excerpt
Temporal introduced Temporal and agentic sandboxes for the OpenAI Agents SDK, combining durable execution with isolated sandbox environments for long-running agents.