OpenAI debuts AgentKit to speed AI agent development

OpenAI on Monday unveiled AgentKit, a suite of developer tools intended to speed the creation, deployment and evaluation of AI agents, announcing the product at its Dev Day event in San Francisco. The toolkit, CEO Sam Altman said, bundles visual design, embeddable chat interfaces, connectors and evaluation tools to move agent projects from prototype to production with less friction.

AgentKit centers on four capabilities. Agent Builder is a visual, drag‑and‑drop canvas for composing agent logic and workflows; Altman compared it to "Canva for building agents" and described AgentKit as "everything you need to build, deploy, and optimize agent workflows with way less friction." ChatKit provides an embeddable chat interface developers can customize and integrate into websites and apps. Evals for Agents expands OpenAI's testing suite with step‑by‑step trace grading, datasets for component assessment, automated prompt optimization and the ability to run evaluations on external models. A Connector Registry lets administrators securely link agents to internal systems and third‑party services while enforcing controls.

OpenAI demonstrated the toolkit onstage, where an engineer, Christina...