OpenAI has released AgentKit, a comprehensive toolkit designed to streamline the development, deployment, and optimisation of AI agents. The toolkit aims to simplify the process of moving AI agents from prototype to production. AgentKit includes Agent Builder, a visual canvas for composing logic using drag-and-drop nodes, connecting tools, and configuring custom guardrails. It also has a Connector Registry, a central place for admins to manage how data and tools connect across OpenAI products, and ChatKit, a toolkit for embedding customisable chat-based agent experiences.
AgentKit expands evaluation capabilities with features such as datasets, trace grading, automated prompt optimisation, and third-party model support. These tools provide developers with a comprehensive view of their agents' performance and areas for improvement. AgentKit builds upon the Responses API, offering a more efficient and reliable way to construct agents. The Agent Builder supports preview runs, inline evaluation configuration, and versioning for rapid iteration. OpenAI engineer Christina Huang demonstrated the simplicity of AgentKit by building an entire AI workflow and two agents onstage in under eight minutes.
AgentKit targets individual developers and enterprises, providing scalable AI solutions. It gives OpenAI a fresh edge over other platforms trying to do the same thing. The toolkit's features aim to address the complexities of building AI agents, such as orchestration, evaluation loops, tool connections and user interface development. AgentKit allows users to build high-quality AI agents tailored to any industry vertical using an intuitive visual builder, evaluation tools, and guardrails.
