OpenAI Introduces New Developer Tools to Build Custom AI Agents
OpenAI launches a new Responses API that provides developers with the underlying components to create autonomous AI agents for business tasks. This new toolkit replaces the older Assistants API and aims to deliver on the growing industry hype surrounding agent technology.
OpenAI releases a new suite of tools designed to help businesses and developers build custom AI agents through its new Responses API. This toolkit allows companies to create automated systems that perform web searches, scan internal files, and navigate websites independently. The Responses API effectively replaces the older Assistants API, which OpenAI plans to officially retire in the first half of 2026.
The launch comes at a time when the hype around AI agents far outpaces actual usable products in the tech industry. OpenAI's API product head Olivier Godement acknowledges that while demonstrating an agent is easy, scaling it for regular use presents a significant challenge. Earlier attempts by OpenAI, such as the Operator and deep research features in ChatGPT, provide a glimpse into this technology but still lack true autonomy.
To power these new custom agents, the Responses API gives developers access to preview versions of GPT-4o search and GPT-4o mini search. OpenAI claims these specialized models offer high factual accuracy, scoring 90% and 88% respectively on the company's SimpleQA benchmark. By selling access to these underlying components, OpenAI hopes developers build applications that finally deliver on the bold promises of the AI agent movement.