OpenAI CEO Sam Altman announced the launch of AgentKit at the company’s Dev Day event. AgentKit is a toolkit designed for building and deploying AI agents. Altman described it as a complete set of building blocks available in the OpenAI platform to help developers take agents from prototype to production. He stated that it provides everything needed to build, deploy, and optimize agent workflows with significantly less friction.
This launch underscores OpenAI’s effort to increase developer adoption by making the process of building agents faster and more straightforward. It also represents a competitive move against other AI platforms that are also racing to offer integrated tools for creating autonomous agents. These enterprise agents are designed to perform complex tasks, moving beyond simply responding to prompts.
AgentKit was one of several announcements made at Dev Day, which also included the new ability to build apps directly inside ChatGPT. The chatbot now has 800 million weekly active users.
AgentKit includes several core capabilities. The first is Agent Builder, which Altman compared to Canva for building agents. He described it as a fast, visual way to design the logic and steps for an agent. It is built on top of the Responses API that hundreds of thousands of developers already use.
The second capability is ChatKit. This provides a simple, embeddable chat interface that developers can use to integrate chat experiences into their own applications. Altman noted that developers can bring their own brand and unique workflows into the product.
Another component is Evals for Agents. This introduces tools to measure AI agent performance. It includes step-by-step trace grading, datasets for assessing individual agent components, automated prompt optimization, and the ability to run evaluations on external models directly from the OpenAI platform.
Finally, AgentKit includes access to OpenAI’s connector registry. This allows developers to securely connect their agents to internal tools and third-party systems through an admin control panel, all while maintaining security and control.
To demonstrate how easy AgentKit is to use, OpenAI engineer Christina Huang built an entire AI workflow and two AI agents live onstage in under eight minutes. Altman commented that this toolkit includes all the features the company wished it had when building its first agents. He also noted that OpenAI has already signed several launch partners who have successfully scaled their agents using AgentKit.