Agentic AI is the technology industry’s latest success story, with companies like OpenAI and Anthropic racing to give businesses the tools they need to create these automated little helpers. To this end, OpenAI has updated its Agent Software Development Toolkit (SDK) to introduce a number of new features designed to help companies create their own agents that run on OpenAI’s models.
New features in the SDK include a sandbox feature that allows agents to operate in a controlled computer environment. This is important because running agents in a completely unsupervised manner can be dangerous due to their sometimes unpredictable nature.
Sandbox integration allows agents to work in siled capacities within specific workspaces, accessing files and code only for specific operations and otherwise protecting the integrity of the entire system.
In this context, the new version of the SDK will also provide developers with a distribution harness for Frontier models, allowing agents to work with files and approved tools within their workspaces, the company said. (In agent development, “harness” is a term that refers to other components of an agent other than the model on which the agent is running. Companies often use distribution harnesses to both deploy and test agents that run on the Frontier model. The Frontier model is considered the most advanced general-purpose model available.)
“The core of this release is to take the existing Agents SDK and make 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-term agents using our harnesses and whatever infrastructure they have on hand,” he said.
These “long-term” tasks are generally considered to be more complex, multi-step tasks.
tech crunch event
San Francisco, California
|
October 13-15, 2026
OpenAI said it will continue to expand the Agents SDK, but initially the new harness and sandbox functionality will be released in Python first, with TypeScript support planned for a later release. The company said it is also working on bringing more agent features to both Python and TypeScript, such as code modes and subagents.
New Agent SDK functionality is available to all customers via API and uses standard pricing.
