Developer Summit ’25 Day 2 Session 2

As engineers, it is hard (and arguably unwise) to trust "magic". What then should we do when our industry is flooded with "magical" new tools that promise to change the game? The best way to see through a magic trick is to learn to do it yourself! Based on the seminal blog post by Thorsten Ball from Amp, "How to Build an Agent, or: The Emperor Has No Clothes", this talk will dispel the illusion of magic surrounding agent-powered code editors like Amp, Claude Code, Copilot, Cursor, Windsurf, and Zed. Together, we will build our own coding agent in only ~400 lines of not-very-exciting JavaScript, and use it to complete some realistic coding tasks. Armed with this clear understanding of how coding agents work under the hood, you will be better able to reason about how and when to use them in your own work.

AI agents are already connecting to frontend development workflows through Model Context Protocol, and the results are impressive. From React component generation to Figma-to-code pipelines, there's a growing ecosystem of MCP servers that frontend developers can plug into today. I've been exploring what's actually working in practice - which integrations deliver real value, what patterns are emerging across successful MCP servers, and where the biggest opportunities still lie. I'll demo some of the most compelling examples, show you how to evaluate and integrate existing MCP tools into your workflow, and share what I've learned about building frontend-focused MCP integrations that developers actually want to use.

Generative AI is great at producing text — but what happens when it can actually do things? This talk explores the shift from AI as a conversational partner to AI as an active participant in real-world workflows. We’ll look at how to connect AI systems to external tools and APIs using the Model Context Protocol (MCP). Instead of stopping at answers, AI can now execute file operations, manage digital assets, analyse images, and automate tasks that normally require human intervention. The result is a move from “chatting with AI” to “working with AI.” Through practical examples, you’ll see how to: <ul> <li>Extend AI beyond text into action-oriented workflows.</li> <li>Safely connect AI models to existing developer tools and services.</li> <li>Combine structured function calls with unstructured conversation.</li> </ul> Build AI-powered assistants that handle files, images, and APIs as part of a single seamless interaction. This session isn’t about abstract futures — it’s about what you can build today. Whether you’re a developer, architect, or product builder, you’ll leave with a clear understanding of how to bridge the gap between natural language input and real-world execution. AI that acts changes the role it plays: from passive responder to proactive teammate. Come see what happens when AI stops talking — and starts doing.