OpenAI updates its Agents SDK to help enterprises build safer, more capable agents

Agentic AI is the tech industry’s latest success story, and companies like OpenAI and Anthropic are racing to give businesses the tools they need to create these automated little helpers. For this purpose, OpenAI has now updated its Agent Software Development Toolkit (SDK)introducing a number of new features designed to help enterprises build their own agents that run on the back of OpenAI models.

New SDK features include a sandboxing capability, which allows agents to operate in controlled computer environments. This is important because agents running completely unattended can be dangerous due to their occasionally unpredictable nature.

By incorporating sandboxing, agents can work in a silent capacity in a specific workspace, accessing files and code only for specific operations, while otherwise protecting the overall integrity of the system.

Relatedly, the new version of the SDK also provides developers with an in-distribution harness for boundary models that will allow those agents to work with files and approved tools in a workspace, the company said. (In agent development, “harness” is a term that refers to the components of an agent other than the model in which it runs. A harness within a distribution often allows companies to develop and test agents running on boundary models, which they are considered to be the most advanced general purpose models available.)

“This release, at its core, is about taking the existing Agents SDK and making it compatible with all these sandbox providers,” Karan Sharma, who works on OpenAI’s product team, told TechCrunch.

The hope is that this, combined with the new harness capabilities, will allow users to “build these long-horizon agents using our harness and with whatever infrastructure they have,” he said.

Such “long horizon” tasks are generally considered more complex and multi-step tasks.

OpenAI said it will continue to expand the Agents SDK over time, but initially, the new braid and sandbox features are being released in Python first, with TypeScript support planned for a later release. The company said it is also working to offer more agent capabilities, such as code mode and subagents, in both Python and TypeScript.

The new Agents SDK features are offered to all customers via the API and will use standard pricing.