š„ļø Beyond Chat: Building Interactive AI Experiences with MCP Apps on AWS
AI assistants are moving beyond text responses. Explore how MCP Apps can bring interactive interfaces into AI experiences and how Amazon Bedrock AgentCore provides a secure, serverless foundation for building these next-generation applications.
š Introduction
What if an AI assistant could do more than send you a text response?
Imagine asking an AI:
"Show me my AWS cost analysis."
Instead of receiving only a paragraph of text, the AI could display an interactive dashboard with charts, filters, buttons, and useful controls directly inside the AI experience.
This is the idea behind MCP Apps.
MCP Apps extend the Model Context Protocol (MCP) to support interactive user interfaces that can be rendered directly inside compatible AI hosts. AWS recently demonstrated how MCP Apps can be built using Amazon Bedrock AgentCore.
š§ What Are MCP Apps?
Model Context Protocol (MCP) provides a standardized way for AI applications to discover and use tools and resources.
Traditionally, an AI tool might return something like:
1
2
3
4
5
AWS Cost:
$128.50
Highest spending service:
Amazon EC2That's useful, but it is still just text.
With an interactive MCP App, the experience could become:
1
2
3
4
5
6
7
8
9
10
11
12
13
āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā
ā AWS Cost Dashboard ā
āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā¤
ā ā
ā Total Cost: $128.50 ā
ā ā
ā EC2 āāāāāāāāāā $72 ā
ā S3 āāāā $25 ā
ā Lambda āā $12 ā
ā ā
ā [View Details] [Filter] ā
ā ā
āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāThe AI experience becomes more interactive rather than being limited to text.
š¬ From Chat Responses to Interactive Experiences
We can think about the evolution like this:
1
2
3
4
5
6
7
8
9
10
11
12
13
Traditional Chatbot
ā
Text Response
ā
AI Assistant
ā
Tool Calling
ā
MCP
ā
MCP Apps
ā
Interactive AI ExperienceThe major change is that AI can become a gateway to interactive applications, not just a conversational interface.
š How MCP Apps Work
A simplified architecture looks like:
1
2
3
4
5
6
7
8
9
10
11
12
13
User
ā
AI Host
ā
AI Agent
ā
MCP Client
ā
MCP Server / App
ā ā
Tool/Data UI
ā ā
Result Interactive ViewThe AI agent can use MCP to interact with tools while the MCP App provides an interface for presenting information or collecting user interaction.
AWS describes MCP Apps as an extension of MCP that enables interactive HTML widgets to be rendered inside AI hosts.
āļø Where Does Amazon Bedrock AgentCore Fit?
Amazon Bedrock AgentCore provides capabilities for building, connecting, and operating AI agents.
For MCP Apps, AgentCore Runtime can provide a secure, serverless, session-isolated environment with native MCP support.
A conceptual architecture could look like:
1
2
3
4
5
6
7
8
9
10
11
12
13
User
ā
AI Application
ā
Agent
ā
AgentCore Runtime
ā
MCP Application
ā ā
Tools UI
ā ā
AWS Data Interactive ViewThis allows developers to focus on the application experience instead of building all the underlying infrastructure themselves.
š A Real-World Example
Imagine an AI Cloud Cost Assistant.
A user asks:
"Show me where my AWS spending increased this month."
The agent could:
- Understand the request
- Retrieve relevant cost information
- Analyze the data
- Identify unusual changes
- Display the result using an interactive UI
For example:
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
User
ā
"Why did my AWS cost increase?"
ā
AI Agent
ā
Retrieve Cost Data
ā
Analyze Changes
ā
Identify EC2 Increase
ā
Interactive Dashboard
ā
User explores detailsInstead of asking follow-up questions repeatedly, the user could interact directly with the generated interface.
š ļø Why Interactive UI Matters
Text is excellent for many AI tasks.
But some information is easier to understand visually.
For example:
Text
Your monthly usage increased by 32%. EC2 contributed most of the increase.
Interactive UI
1
2
3
4
5
6
7
8
9
Monthly Usage
January āāāāāāāāāāā
February āāāāāāāāāāāāāāā
March āāāāāāāāāāāāāāāāāā
Increase: +32%
[View EC2] [View S3] [View Lambda]The second approach allows users to explore the information instead of simply reading it.
This can be especially useful for:
- š Dashboards
- š° Cost analysis
- š Analytics
- š Shopping assistants
- šļø Scheduling applications
- š Business workflows
- š§āš» Developer tools
š Security Still Matters
Interactive AI applications need strong security controls.
When an AI agent can access tools and data, developers should carefully control:
- Authentication
- Authorization
- Tool permissions
- Data access
- User sessions
- API access
- Sensitive information
A useful principle is:
The AI should only access the resources required for the user's task.
Amazon Bedrock AgentCore provides capabilities around secure agent execution and identity that can help developers build controlled agentic applications.
ā” Why This Is Interesting for Developers
MCP Apps create an interesting possibility:
One AI conversation can become an entry point to many applications.
Instead of building a separate interface for every task, developers can create tools and interactive experiences that AI applications can discover and use.
This changes how we think about application interfaces.
The interface may no longer be:
1
2
3
4
5
6
7
8
9
Website
ā
Menu
ā
Button
ā
Form
ā
ResultIt could become:
1
2
3
4
5
6
7
8
9
10
11
User
ā
AI
ā
Intent
ā
Tool
ā
Interactive Experience
ā
Actionš The Future of AI Interfaces
AI interfaces are evolving quickly.
We can imagine three stages:
Stage 1 ā Chat
1
User ā Question ā AI ā TextStage 2 ā Tool-Using AI
1
User ā AI ā Tool ā Data ā ResponseStage 3 ā Interactive AI
1
2
3
4
5
6
7
8
9
User
ā
AI
ā
Tools + Data
ā
Interactive UI
ā
User ActionMCP Apps are an interesting step toward the third stage.
šÆ Conclusion
AI applications are no longer limited to generating text.
With MCP, AI agents can interact with tools and resources. With MCP Apps, those interactions can become more visual and interactive.
Amazon Bedrock AgentCore provides a serverless, session-isolated environment for building these types of agentic applications.
The future of AI interfaces may not be about choosing between a website and a chatbot.
It could be about combining both:
Conversation for understanding.
Tools for action.
Interactive UI for experience.
As AI continues to evolve, developers have an opportunity to build applications where the conversation itself becomes the starting point for an interactive digital experience.
š” What would you build as an interactive MCP App? Share your idea in the comments!
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article