OpenAI updates its Agent SDK to help companies create more secure and capable agents

Agentic AI is the tech industry’s latest success story, and companies like OpenAI and Anthropic are racing to give companies the tools they need to create these little automated helpers. To that end, OpenAI has now updated its software development toolkit (SDK) for agentsintroducing a number of new features designed to help companies create their own agents that run on the back of OpenAI models.

New SDK capabilities include a sandbox capability, which allows agents to operate in controlled computing environments. This is important because running agents completely unattended can be risky due to its occasionally unpredictable nature.

With sandbox integration, agents can work in an isolated capacity within a particular workspace, accessing files and code only for particular operations, while protecting the overall integrity of the system.

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

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

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

These “long-horizon” tasks are generally considered to be more complex, multi-step jobs.

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

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