Building products for AI agents first requires API coverage, skills, and MCP infrastructure¶
Insight: According to Peter Yang, the "user" in UX is rapidly becoming an agent. Products must prioritize AI agent access through complete API coverage, skills (text files instructing AI how and when to use APIs), and MCP servers (bundling API connections, authentication, and tool definitions). The hierarchy: "MCPs are what your AI can do. Skills are how it should do it."
Detail: Yang uses a kitchen metaphor to explain the infrastructure layers: APIs are "specific tools" (endpoints), Skills are "recipes" (markdown instructions), and MCPs are "fully equipped kitchens" (bundled servers). The article's first two practical steps are shifting the mindset from human-first to agent-first UX, and ensuring every important product capability has a corresponding API endpoint. Remaining steps were behind a paywall. Yang cites OpenClaw founder Peter Steinberger building 40 command-line tools before his AI product succeeded — demonstrating early conviction in agent accessibility.