San Francisco, October 6, 2025: OpenAI has announced a big new tool for developers. The company launched a product called AgentKit at its annual DevDay conference on Monday, October 6, 2025. This new toolkit is a major step toward making building, deploying, and running AI agents much simpler.
AI agents are smart systems. They can plan steps and perform tasks independently. This is unlike a simple chatbot that just answers questions. Before AgentKit, making these agents was a very complicated process. Developers had to use many different tools and spend many weeks of effort.
OpenAI CEO Sam Altman said that AgentKit gives a “complete set of building blocks.” These blocks help developers move their AI agent projects from a test idea to a finished product. The toolkit aims to bring a lot of value to all users, from individual developers to large companies.
Key Tools in AgentKit
AgentKit bundles all the important parts needed for agent creation. It combines visual design tools, data connection features, and evaluation systems. The three main components are the Agent Builder, ChatKit, and the Connector Registry.
Agent Builder: This is the most important part. It provides a visual canvas, or a ‘no-code’ builder. Developers can design complex, multi-step agent workflows just by dragging and dropping components. It helps teams see exactly how their agents will work. It also has features for version control, so you can track and manage changes easily.
ChatKit: This tool makes it easy to add the AI agent to an app or website. It provides a plug-and-play toolkit to embed a customizable chat interface. This means companies can quickly put their new AI agents to work for things like customer support. The chat interface can be customised to match a company’s brand or website design.
Connector Registry: Agents need data to work. The Connector Registry is a central place for company admins. It helps them manage how their internal data and tools connect securely with the OpenAI platform. This includes popular business services like Dropbox, Google Drive, Microsoft Teams, and SharePoint.
Focus on Safety and Performance
Building reliable AI agents is very important. AgentKit includes key features for quality and security.
Evals: OpenAI has expanded its Evals (evaluation) framework. It includes new features like datasets and trace grading. These features help developers systematically test an agent’s performance. They can automatically grade the agent’s work and check full workflows to make sure they are accurate. It even allows testing models from other companies.
Guardrails: These are a layer of safety controls. They are designed to prevent the AI agents from misusing data or performing harmful actions. For example, they can help in masking sensitive personal information.
Reinforcement Fine-Tuning (RFT): This feature helps train the models to make better decisions and follow specific rules given by the developer. It allows for advanced customisation of the model’s reasoning.
Availability and Impact
OpenAI has made several parts of the toolkit available right away. ChatKit and the new Evals features are now generally available to all developers. The Visual Agent Builder and the Connector Registry are currently in beta (testing phase).
AgentKit runs on the standard OpenAI API pricing. This makes it more accessible for a wider range of developers and businesses. By simplifying the process so much, this launch is expected to speed up how quickly AI agents are adopted in real-world business tasks. Experts say that AgentKit is a significant step towards a future where AI agents become a common part of how companies work.
