Liam Hampton

Beyond plain text: The Model Context Protocol The Model Context Protocol (MCP) provides an open standard for connecting Large Language Models to local and remote data sources. While early iterations of AI chat relied on ASCII art and emojis to simulate visualization, Marlene Mhangami explains that MCP now enables servers to return interactive components. This shifts the LLM from a simple text generator to a dynamic engine capable of rendering complex UI elements. The protocol architecture consists of three pillars: hosts like Visual Studio Code, clients such as GitHub Copilot, and lightweight servers that expose specific tools or resources. Anatomy of the MCP app interaction flow When a user prompts a host with a request like "Show me analytics," the system triggers a multi-step sequence. The GitHub Copilot agent identifies the relevant Model Context Protocol tool and calls the server. Critically, the server does not just return raw data; it provides a UI resource reference pointing to an HTML element. Visual Studio Code then fetches this HTML and renders it within a sandboxed iframe. This sandboxing is vital for security, preventing the third-party UI from accessing editor settings or sensitive APIs. Building a flame graph profiler with Go and React Liam Hampton demonstrates building a real-world application that profiles Go code using `pprof`. The server, written in TypeScript, executes a Go binary running sorting algorithms and captures performance data. On the frontend, a React application uses hooks to ingest this data and render a live flame graph directly in the chat window. ```typescript // Server-side registration of the UI resource server.setRequestHandler(ListResourcesRequestSchema, async () => { return { resources: [{ uri: "mcp://flamegraph/ui", name: "Flamegraph UI", mimeType: "text/html" }] }; }); ``` Industrial applications and security best practices Companies like Shopify and Figma are already adopting MCP apps to maintain brand consistency within AI interfaces. Shopify, for instance, allows users to complete entire checkout flows without leaving the chat. However, Marlene Mhangami warns users to only source servers from trusted repositories like the official Visual Studio Code extensions marketplace to avoid executing malicious code through the protocol.