OpenAI will allow anyone to create AI agents that will control PCs and surf the Internet

OpenAI has unveiled a tool for developers that will make it easier to create AI agents. The key element here is the new Responses API, which is a set of ready-made solutions for automating routine processes such as searching the Internet, analyzing files, and performing tasks on a computer on behalf of the user.

Image source: Zac Wolff / Unsplash

OpenAI API Platform Lead Olivier Godement said the company sees a big future in AI agents. He said OpenAI is ready to develop its own agents, such as Deep Research and Operator, but given the complexity of the modern world and the diversity of technologies, it wants to provide developers with basic tools to create agents for their unique tasks.

Responses API is equipped with a search tool that uses the same AI model as ChatGPT. Based on GPT-4o and GPT-4o mini, the tool allows you to get relevant information from the Internet with sources, as well as perform tasks on the computer through the Operator model. At the same time, the file search function can be useful for absolutely any purpose, for example, for automating the work of the company’s support service or legal analysis of documents.

Additionally, OpenAI introduced the Agents SDK. This is a tool for managing the work of AI agents. OpenAI product manager Nikunj Handa explained that the Responses API performs individual tasks, while the Agents SDK allows you to coordinate the work of several agents so that they can solve more complex problems together. According to the company, this should simplify the process of managing agents, and at the same time increase their efficiency.

It should be noted that the innovation will complement existing OpenAI solutions, including the Chat Completions API for responding to user requests. In addition, the company plans to abandon the current Assistants API in favor of the new Responses API system by mid-2026.