OpenAI has launched AgentKit, a new set of tools for developers and businesses to build, deploy, and optimize AI agents. According to the company, the toolkit aims to consolidate a process that previously involved fragmented tools, complex orchestration, and manual work. The launch signals a competitive move to make agent building faster and easier for a less technical audience.
AgentKit consists of several core components designed to work together.
- Agent Builder: A visual canvas where users can design and version agent workflows using a drag-and-drop interface. OpenAI CEO Sam Altman described it as being “like Canva for building agents.”
- ChatKit: A toolkit that allows developers to embed customizable chat-based interfaces into their own applications or websites.
- Connector Registry: A central control panel for administrators to manage how agents securely connect to data sources and third-party systems like Google Drive or Microsoft Teams.
The toolkit also expands OpenAI’s evaluation capabilities with new features to measure agent performance, such as trace grading, automated prompt optimization, and support for testing models from other providers.
According to OpenAI, ChatKit and the new evaluation features are generally available. Agent Builder is in beta, and the Connector Registry is rolling out in beta to select enterprise customers. The tools are included with standard API model pricing.
Additional source: TechCrunch