OpenAI today introduced AgentKit, a new suite of software tools designed to help companies build and deploy AI “agents” without the heavy custom engineering that has traditionally slowed adoption. The release marks one of the company’s largest pushes yet into enterprise software, signaling a shift from model development to full-stack agent deployment infrastructure.
AgentKit combines multiple components that streamline the creation of automated AI workflows. The new Agent Builder gives developers a visual canvas to design and modify multi-agent systems, allowing teams to drag, connect, and version logic flows rather than write complex orchestration code. The interface supports live previews, in-line evaluation, and version control, reducing the time from prototype to deployment.
Enterprises can also manage data connections through the Connector Registry, a central control panel for connecting internal systems such as Microsoft Teams, SharePoint, Dropbox, Google Drive and third-party Model Context Protocol (MCP) tools. OpenAI says the registry is intended to give administrators tighter control over data access and governance across ChatGPT and API-driven workspaces.
To simplify embedding agents into customer-facing products and internal tools, OpenAI launched ChatKit, which allows developers to drop customizable agent chat interfaces directly into applications or websites. Companies like Canva and HubSpot are already using it for customer support agents.
OpenAI is also expanding agent evaluation capabilities with new dataset tools, automated grading, and optimization features. The platform now allows developers to test workflows end-to-end, identify weaknesses, and automatically refine prompts. The evaluation system supports both OpenAI models and third-party models.
The company additionally highlighted reinforcement fine-tuning options that let enterprises train agents to make better decisions, including how and when to call external tools. GPT-5 is currently in private beta for fine-tuning use cases.
Agent Builder is available in beta, while ChatKit and the new evaluation tools are available to all developers starting today.
OpenAI says it plans to introduce a standalone Workflows API and new deployment pathways inside ChatGPT in the coming months.
Need Help?
Keeping track of the growing AI regulatory landscape can be difficult. So if you have any questions or concerns, don’t hesitate to reach out to BABL AI. Their Audit Experts can offer valuable insight, and ensure you’re informed and compliant.
