OpenAI launches AgentKit to help developers build and ship AI agents

In an exciting announcement during the recent Dev Day event, OpenAI CEO Sam Altman introduced AgentKit, a comprehensive toolkit designed to facilitate the development and deployment of AI agents. "AgentKit is a full suite of resources available on the open AI platform, aimed at helping developers transition their agents from initial concept to full-scale production," Altman explained. This endeavor is part of OpenAI's broader strategy to enhance developer engagement by streamlining the agent-building process. The launch of AgentKit signifies OpenAI's commitment to staying competitive in a rapidly evolving landscape, where various AI platforms are striving to provide integrated solutions for creating autonomous agents capable of handling complex tasks. One of the standout features unveiled at the event was the ability to construct applications directly within ChatGPT, which has recently achieved a staggering 800 million weekly active users. AgentKit boasts several key functionalities, starting with the Agent Builder, which Altman likened to Canva for agent creation. "It offers a quick, visual approach to designing the logic and steps involved in building agents," he said, emphasizing that it utilizes the responses API already familiar to countless developers. Another pivotal component is ChatKit, which supplies an easy-to-integrate chat interface that developers can embed into their own applications. This allows for a customized experience that reflects each brand's unique identity and workflows. Additionally, AgentKit introduces Evals for Agents, a set of tools designed to evaluate AI agent performance. This includes detailed grading systems, datasets for evaluating specific agent components, and automated optimization of prompts, along with the ability to conduct assessments on external models directly via the OpenAI platform. Moreover, AgentKit provides access to OpenAI's connector registry, enabling developers to securely integrate agents with internal tools and third-party systems through an intuitive admin control panel, ensuring security and oversight throughout the process. To demonstrate the accessibility of AgentKit, OpenAI engineer Christina Huang successfully built an entire AI workflow and two AI agents in a live presentation lasting under eight minutes. "This encompasses everything we wished we had when we were developing our first agents," Altman remarked, revealing that several partners have already begun scaling their agents with the new toolkit.