To build an AI agent, developers currently have to manage complex orchestration, create custom connectors, build evaluation pipelines, and handle several other critical tasks. At DevDay 2025, OpenAI announced AgentKit, a new set of tools that make it easier for developers to build, deploy, and optimize AI agents using OpenAI’s models.
OpenAI’s AgentKit includes three main components:
- Agent Builder - a WYSIWYG interface for creating and versioning multi-agent workflows.
- Connector Registry - a unified location for admins to manage tools and data.
- ChatKit - a toolkit that enables developers to easily embed customizable, chat-based agent experiences into any product.
As shown in the screenshot below, AgentKit’s Agent Builder allows developers to create agents using drag-and-drop nodes. Developers can also configure guardrails, run evaluations of their agents, and preview runs within this new visual canvas. Guardrails are an open-source, modular safety layer that protects agents from malicious behavior, for example, by detecting PII, jailbreak attempts, and other risks.
Creating an agent is only part of the solution. The agent should be deployed in an easy-to-use way for consumers. Deploying agents with chat UIs for end-users involves handling streaming responses, managing multiple chat threads, designing in-chat experiences, and more.
The new ChatKit will allow developers to easily embed chat-based agents into any app or website. It also allows developers to customize the UI to match their desired theme or brand.
Finally, OpenAI announced the following improvements to Evals to make it easier for developers to test their agents:
- Datasets–rapidly build agent evals from scratch and expand them over time with automated graders and human annotations..
- Trace grading–run end-to-end assessments of agentic workflows and automate grading to pinpoint shortcomings.
- Automated prompt optimization–generate improved prompts based on human annotations and grader outputs.
- Third-party model support–evaluate models from other providers within the OpenAI Evals platform.
Agent Builder is now available in beta for all developers. Connector Registry is now rolling out in beta to select API, ChatGPT Enterprise, and Edu customers with a Global Admin Console. ChatKit and the new Evals improvements are generally available to all developers from today.