OpenAI launches AgentKit to help developers build and ship AI agents (techcrunch.com)

OpenAI announced AgentKit at Dev Day — a bundled toolkit intended to take AI agents from prototype to production with less friction. Key components include Agent Builder, a visual “Canva-like” interface built on OpenAI’s Responses API for designing agent logic and workflows; ChatKit, an embeddable, brandable chat UI; Evals for Agents, which brings step-by-step trace grading, component-level datasets, automated prompt optimization, and the ability to run evaluations on external models directly from the OpenAI platform; and a connector registry plus admin control panel for securely wiring agents to internal tools and third-party systems. OpenAI demonstrated rapid prototyping live, building a full workflow and two agents in under eight minutes, and says several launch partners have already scaled agents using the toolkit. For the AI/ML community, AgentKit lowers the engineering and operational barriers to building autonomous, multi-step agents and standardizes evaluation and security practices. The built-in evals and connector controls help teams measure, iterate, and govern agents in production, while support for external-model evaluation promotes interoperability. Strategically, it strengthens OpenAI’s developer ecosystem and competes with other platforms offering integrated agent stacks — which may speed enterprise adoption but also raises questions about platform lock‑in and dependency on OpenAI tooling and connectors.