On this episode of the Modern Web Podcast, hosts Rob Ocel, Danny Thompson, and Adam Rackis are joined by Tejas Kumar, host of The Contagious Code podcast, author of Fluent React, and Developer Relations Engineer for Generative AI at DataStax. They unpack the current wave of AI announcements from Google I/O and Microsoft Build, and zoom in on the significance of MCP (Model Context Protocol) as a foundational shift in how AI-powered apps will be built and used.

Tejas breaks down what MCP is, why it's catching on across the industry, and how it could become the HTTP of AI apps. The group explores real-world examples, like AI apps managing your inbox or booking flights without ever opening a browser, and discuss how MCP servers enable secure, agent-driven experiences that can act on your behalf. They also touch on hallucinations, the role of fine-tuning vs. tool integration, and the future of checkout flows powered by AI agents.

Keypoints from this Episode:

- MCP enables structured communication between AI apps and servers, allowing agents to perform real tasks like sending emails or booking flights

- Users will increasingly interact with applications through natural language, with agents handling workflows behind the scenes

- Connecting models to tools via MCP helps reduce hallucinations by ensuring actions and responses are grounded in real data

- Most use cases benefit more from retrieval-augmented generation and strong tool integration than from expensive model fine-tuning

Follow Tejas on Social Media

Twitter: https://x.com/TejasKumar_

Linkedin: https://www.linkedin.com/in/tejasq/