OpenAI Upgrades Agents SDK With Safer Sandbox Tools

OpenAI has updated its Agents SDK with new features aimed at helping enterprises build more capable and safer AI agents. Key additions include sandboxing, which lets agents operate in controlled environments to limit unpredictable behavior, and an in-distribution harness for frontier models that enables deployment and testing within approved workspaces. The updates are designed to support complex, long-horizon tasks while keeping systems secure. Initially launching in Python, with TypeScript support planned later, the new capabilities are available to all API customers at standard pricing. OpenAI says further agent features, including code mode and subagents, are also in development.