OpenAI's new tools help developers build computer-operating AI agents
What's the story
OpenAI has launched a new Responses API, designed to help developers build their own AI agents. These agents would be able to search the web, dig through files, and perform tasks on a computer. Olivier Godement, OpenAI's head of product for the platform, said he was excited about giving developers the foundational tools to build custom agents for their own needs.
API advantages
Responses API features web search and document scanning tools
The Responses API offers a web search tool built on the same model ChatGPT uses for search. This lets developers fetch real-time information and citations from the internet while using GPT-4o and GPT-4o mini. It also provides a computer-use feature that employs OpenAI's Operator model to perform tasks on behalf of users. Another handy feature is its capability to sift through large volumes of documents, which can be useful for customer support agents or legal assistants.
SDK introduction
OpenAI launches agents SDK for AI agent workflow management
Along with the Responses API, OpenAI has also announced the Agents SDK. This tool will help developers orchestrate the workflow of their AI agents seamlessly. Nikunj Handa, a product manager for OpenAI's API team, described it as a way for multiple atomic units to work together to solve complex tasks. This will simplify agent management and ensure they all work toward a common goal.