This feature gives developers a drag-and-drop canvas to design multi-step agent workflows with nodes, branching logic, tool integrations and versioning. It simplifies building complex agent logic by making the flow visual rather than purely code-based.

ChatKit provides a ready-to-use, customizable chat interface that you can embed into your application or website, enabling users to interact with agents built via AgentKit without building the chat layer from scratch.

AgentKit includes a centralized registry to manage connections between agents and external systems — e.g., data sources, APIs, internal tools — allowing secure, controlled tool usage and simplifying orchestration of real-world workflows.

This feature enables developers to assess agent performance using datasets, trace grading, prompt optimisation and support for third-party models. It gives a feedback loop so you can measure and improve agent accuracy, behaviour and reliability.

AgentKit supports versioning of workflows, preview and publish capabilities, enabling a shift from prototype to production with better governance, rollback control, and monitoring of agent behavior in live environments.