<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>REST on</title><link>https://eunus.dev/tags/rest/</link><description>Recent content in REST on</description><generator>Hugo -- gohugo.io</generator><lastBuildDate>Sun, 14 Jun 2026 00:00:00 +0600</lastBuildDate><atom:link href="https://eunus.dev/tags/rest/index.xml" rel="self" type="application/rss+xml"/><item><title>What Building an MCP Server Taught Me About API Design</title><link>https://eunus.dev/blog/what-building-an-mcp-server-taught-me-about-api-design/</link><pubDate>Sun, 14 Jun 2026 00:00:00 +0600</pubDate><guid>https://eunus.dev/blog/what-building-an-mcp-server-taught-me-about-api-design/</guid><description>The task seemed straightforward: build an MCP (Model Context Protocol) server on top of an existing production system so an AI agent could interact with it. I was the one writing the MCP layer — designing the tools, mapping them to the existing API endpoints, and making the backend legible to an AI. The backend was live. The APIs were working. A frontend client was consuming them without complaints.
I was working under strict constraints — no changes to the backend, no changes to the existing client.</description></item></channel></rss>