Training > AI/Machine Learning > MCP Apps Development (RXM405)
INSTRUCTOR-LED COURSE

MCP Apps Development (RXM405)

AI agents are increasingly expected to do more than return text and structured data. MCP Apps extend the Model Context Protocol with interactive interfaces, letting hosts present dashboards, forms, and review flows directly to users. This extension is becoming central to how developers build rich human-agent collaboration.

Who is it for?

For developers, AI and platform engineers, architects, and technical leads who want to build and secure interactive MCP Apps for production agentic workflows.
read less read more
What you'll learn:

Design MCP tools that return app views and build resources that render safely inside hosts. Connect apps through the app bridge and secure them against prompt injection and unsafe actions.
read less read more
What it prepares you for:

Prepare to take on specialized roles such as AI Platform Engineer, MCP Developer, and AI Solutions Architect. Architect and deploy interactive MCP Apps within enterprise AI systems.
read less read more
Course Outline
Expand All
Collapse All
Course Introduction
MCP Apps Overview and Architecture
- MCP Apps as an official extension for interactive UI in MCP-capable hosts
- Relationship between MCP servers, tools, resources, hosts, app views, and users
- How app views differ from text, Markdown, structured data, RAG output, and traditional web apps
- Common use cases for dashboards, forms, review flows, media, maps, reports, and operational tools
- Host compatibility, lifecycle, trust boundaries, and production adoption considerations
- LAB: Map an MCP App use case to tools, resources, views, and host responsibilities
Building MCP App Tools and Resources
- Designing tools that return structured results and app-capable responses
- Serving app resources with the required metadata and content types
- Connecting app views to tool outputs, resources, prompts, and server capabilities
- Handling result refresh, state, errors, loading states, and partial data
- Packaging app code, assets, schemas, and server implementation for repeatable use
- LAB: Build an MCP server tool that returns an interactive app view
App Bridge, Interaction, and UX Patterns
- App-to-host communication, bridge APIs, and message flow
- Calling MCP tools from an app view and receiving fresh server results
- Designing forms, filters, drilldowns, visualizations, and action review screens
- Keeping users oriented with context, status, permissions, confidence, and provenance
- Human-in-the-loop confirmation for writes, external actions, and side effects
- LAB: Add interaction and server-backed updates to an MCP App view
Security, Testing, and Production Readiness
- Sandboxing, content security, untrusted content, and host rendering constraints
- Permission boundaries for tools, resources, user data, and external service access
- Protecting against prompt injection, data leakage, confused deputy risks, and unsafe actions
- Debugging app rendering, bridge communication, tool calls, schemas, and host compatibility
- Operational practices for versioning, observability, accessibility, and rollout
- LAB: Test, debug, and harden an MCP App before production use

Prerequisites
Programming experience, preferably with JavaScript or TypeScript. Basic familiarity with MCP concepts such as clients, servers, tools, resources, and prompts. Experience with web UI development is helpful but not required.
Lab Info
An Ubuntu cloud instance is provided. Students need the ability to SSH or browse to the cloud instance.
About this Course
The Linux Foundation has partnered with the AI/ML experts at rx-m to offer this course to the community.