AWS demos generative UI for AI agents on Bedrock AgentCore with AG-UI

The AWS Machine Learning Blog has published a walkthrough of how the AG-UI protocol integrates with the Fullstack AgentCore Solution Template (FAST) to build interactive frontends for AI agents on Amazon Bedrock AgentCore. CopilotKit adds generative UI and human-in-the-loop support on top.

What is Amazon Bedrock AgentCore

Amazon Bedrock AgentCore is an AWS service for deploying and managing AI agents in production. It handles infrastructure work such as scaling, agent memory, and tool-call orchestration, so developers can focus on the agent's own logic. The Fullstack AgentCore Solution Template (FAST) is a ready-made template for quickly launching a complete application on AgentCore, frontend included.

How AG-UI and CopilotKit work together

AG-UI is an open protocol that standardizes how events pass between an AI agent and the user interface. Instead of only replying with text, the agent can update the UI state in real time. In the post, AWS shows how this stack fits into FAST.

• The AG-UI protocol connects to the Fullstack AgentCore Solution Template (FAST)
• CopilotKit adds generative UI, shared state, and human-in-the-loop
• The entire solution deploys on Amazon Bedrock AgentCore

CopilotKit extends the base integration in two ways. The agent can go beyond displaying data and dynamically generate UI elements tailored to a specific task. And a human can step into the agent's workflow before it carries out a critical action. That is a key pattern for agents trusted with real operations rather than just advice.

What it means

Generative UI and human-in-the-loop are becoming standard building blocks for production agents on AWS. The platform is clearly moving away from purely chatbot-style AI agents toward full interactive applications with visual interfaces that the agent assembles itself for each task.