⚡️The new OpenAI Agents Platform

Key takeaways

OpenAI has launched significant updates to its API, including the Responses API, which integrates new tools and supports a wider range of applications.
The web search tool is designed to improve accuracy and usability, providing developers with real-time access and citations for enhanced data retrieval.
Computer use tools empower developers to automate and execute tasks on computer interfaces, marking a new frontier in AI-driven user assistance.
The Agents SDK has been revamped to provide greater oversight and flexibility in multi-agent orchestration, incorporating feedback from the developer community.
OpenAI assures continued support for chat completions while promoting the new Responses API as the preferred foundation for future developments.

Summary

Launch of New OpenAI Tools

OpenAI announced a series of exciting new developments including the Responses API, a unified API designed to support various agentic applications. This API is a superset of previous chat completion tools, allowing for more complex interactions and integration of built-in tools, such as web search, file search, and a computer use tool. The improvements focus on enhancing the overall developer experience while maintaining access to the existing chat completions API.

Responses API Explained

The Responses API aims to address developer concerns regarding API transitions, particularly around the longevity of chat completion functionalities. OpenAI confirmed that chat completions will continue to be supported for years, while the new Responses API is being positioned as the ideal starting point for both new users and existing developers looking to leverage additional capabilities.

Web Search Tool Capabilities

The web search tool boasts impressive accuracy and new functionality compared to its predecessors, enabling developers to retrieve real-time information with inline citations. There are two access methods: via the Responses API and through the chat completions model itself. The focus is on combining various tools for enhanced functionality and structured outputs tailored for applications.

Integration of Computer Use Tools

The introduction of computer use tools allows developers to build agents capable of automating tasks on users' computers or browsers. This tool can execute clicks, scrolls, and typing actions based on visual inputs, facilitating multi-step processes for users. The exploration of its capabilities includes possible benchmarks in gaming, such as playing Pokémon.

Overview of the Agents SDK

The upgraded Agents SDK includes improvements such as tracing and observability, designed to aid developers in creating and managing multi-agent workflows more effectively. It allows the separation of logic between agents and introduces new features that enhance the user experience for managing these workflows.

Show notes

While everyone is now repeating that 2025 is the “Year of the Agent”, OpenAI is heads down building towards it. In the first 2 months of the year they released Operator and Deep Research (arguably the most successful agent archetype so far), and today ...