OpenAI has launched AgentKit, a suite of tools aimed at accelerating the creation and management of AI agents while enhancing safety. This announcement, made on October 6, 2025, during OpenAI’s Developers’ Day, marks a major advancement in simplifying AI development for both businesses and individual developers. Previously, developing AI agents was an intricate and lengthy task that required juggling various tools, writing custom code, and spending weeks crafting functional workflows. AgentKit transforms this process by offering a consolidated platform that allows developers to visually design workflows, test AI behavior, and deploy chat-based agents more easily.
A key feature, Agent Builder, serves as a visual workflow tool that lets developers drag and drop components to build multi-agent workflows, enabling teams to visualize agent operations, test their performance, and implement iterative improvements swiftly. To facilitate seamless data management, OpenAI has also introduced the Connector Registry, which integrates with widely used applications like Dropbox, Google Drive, Microsoft Teams, and SharePoint, thus improving data accessibility across various workspaces. Furthermore, AgentKit includes Guardrails, a security feature aimed at preventing agents from executing harmful actions or disclosing sensitive information. For chat functionalities, ChatKit allows developers to easily embed AI agents into websites and applications.
These chat agents can fulfill diverse roles such as customer service, onboarding, research, or serving as knowledge assistants. The new Evals features significantly enhance testing and evaluation, enabling developers to automatically grade agents, review entire workflows, and even assess models from external providers. Large organizations like Carlyle and Box have reported quicker development cycles and increased accuracy due to these tools. Additionally, the introduction of reinforcement fine-tuning (RFT) enables models to adhere to custom rules and make improved decisions, with RFT currently available for some models and in beta for GPT-5. ChatKit and Evals are available to all developers, while Agent Builder and Connector Registry are in beta testing.
All tools come under standard API pricing, with OpenAI promising future improvements, including a Workflows API and more deployment options. Through these features, AgentKit seeks to eliminate the challenges that have previously hindered AI agent development, providing a comprehensive, secure, and user-friendly platform for developers and businesses eager to innovate in the AI sector.