In this AI Tech Experts Webinar, Michał Pstrąg, ML Engineer, introduces MCP Apps — a new standard that enables embedding fully interactive UIs directly inside AI chat interfaces.
Instead of redirecting users to external dashboards or SaaS platforms, MCP Apps allow teams to ship usable functionality where users already are — inside the conversation. This shift changes how products are discovered, how conversions happen, and how AI-native experiences are designed.
The session covers:
- how MCP Apps extend the Model Context Protocol
- the architectural flow: server, host, view and tool invocation
- agent-driven UX vs traditional search-based discovery
- practical development guidelines and SDK usage
- evaluation patterns using Promptfoo to avoid brittle behavior
- when MCP Apps make sense — and when a classic web app is still better
If you’re building AI agents, LLM applications, or conversational interfaces in production, this session gives both architectural clarity and practical implementation advice.
If you have questions for Michal, feel free to ask them in the comments and continue the discussion there
Timeline
00:00 Intro & Agenda
01:11 What MCP Apps are and why they matter
02:08 From MCP to industry standard adoption
03:28 Agent-driven user conversion model
05:57 Practical use cases of MCP Apps
07:28 MCP architecture and invocation flow
11:24 Evals, development tips
13:07 Final recommendations
Speaker
Michał Pstrąg
ML Engineer at deepsense.ai
