Last January, Sam Altman, co-founder and CEO of OpenAI, stated: "We believe that by 2025, we could see the first AI agents 'joining the workforce' and materially changing company production." After presenting its own AI agents, Deep Research and Operator, the company introduces a new version of the API Responses and the open-source toolkit SDK Agents, aiming to facilitate developers and companies in creating AI agents tailored to their needs.
A Toolset Simplifying Agent Creation
By mid-2026, OpenAI plans to gradually replace its API Assistants with API Responses to offer developers a more flexible and efficient solution for integrating AI agents into existing systems.
This new version of the API includes tools such as web search (powered by GPT-4o and GPT-4o mini models, similar to search in ChatGPT), file search, and computer usage, thus facilitating real-time data exploitation and the automation of tasks previously hard to mechanize. The computer usage tool, based on Operator's CUA (Computer Usage Agent) model, will allow AI agents to interact directly with business software user interfaces, bypassing the limitations of systems without accessible APIs. This feature opens up new perspectives, notably for automating manual processes still prevalent in many sectors.
Large-Scale Workflow Automation
In addition to the API Responses, OpenAI introduces the SDK Agents, an open-source toolkit designed to orchestrate complex workflows involving multiple AI agents. With this SDK, companies can define specialized agents, assign them distinct roles, and enhance information exchange between them. This framework facilitates the optimization of various processes such as content generation, code review, customer support, document analysis, or business relationship management.
The integration of safeguards, such as security validations and traceability tools, aims to ensure controlled adoption and minimize operational errors. Furthermore, OpenAI relies on a modular approach, allowing developers to adapt these tools to the specific needs of their business.
The company comments:
"With today's releases, we introduce the first building blocks to enable developers and companies to create, deploy, and evolve reliable and high-performing AI agents more easily. As model capabilities become increasingly agentive, we will continue to invest in deeper integrations between our APIs and new tools to help deploy, evaluate, and optimize agents in production. Our goal is to offer developers a seamless platform experience for creating agents that can help accomplish various tasks in any sector."