OpenAI’s Responses API: A New Foundation for Building with Models & Tools

Last month, OpenAI introduced the Responses API: a programmatic agent API businesses can use to perform a wide variety of tasks. With this new primitive, we radically simplified integration, transforming what previously took hundreds of lines of code into just a few. Built from the ground up based on insights from thousands of developers who have used Chat Completions and Assistants APIs, Responses reimagines simplicity, performance, and flexibility, enabling seamless integration of advanced reasoning, multimedia inputs, and multi-step workflows. In this talk, I'll walk through the design decisions and engineering challenges behind Responses. You'll learn how we anticipated developer needs to create an API uniquely engineered for agent-like use cases, capable of handling simultaneous tool calls and seamless multi-turn conversations. We'll explore key features like built-in state management, semantic streaming, intelligent token truncation, and support for hosted tools (ex: web search, file search, and computer operations) that significantly reduce complexity and enhance real-time interactions. And, we’ll talk about how Responses empowers developers to build faster, smarter, and more responsive AI applications than ever before – driving the next wave of intelligent, agentic experiences.