Doppl by Google

Doppl by Google is a research prototype that lets users build and explore interactive AI agents directly in the browser. It provides a visual, node-based interface where you can design agent workflows by connecting components such as large language models, tools, data sources, and custom logic. Users can define how an agent perceives input, reasons through tasks, calls external APIs, and responds, without needing extensive backend infrastructure.

The tool supports rapid iteration: you can test prompts, adjust parameters, and trace the agent’s step-by-step behavior in real time. Doppl includes features for branching logic, memory handling, and multi-step reasoning, making it suitable for experimenting with complex conversational flows or task-oriented agents. It also enables inspection of intermediate outputs, helping users understand why the agent behaves a certain way and refine its design accordingly.