Introducing Quarkus Agent MCP: teaching AI agents to speak Quarkus

Quarkus Agent MCP is a standalone MCP (Model Context Protocol) server that gives AI coding agents like Claude Code, Cursor, and GitHub Copilot deep Quarkus-specific context. It runs as a separate process wrapping `quarkus dev`, so it stays alive even when the application crashes, allowing agents to read logs, inspect exceptions, and fix code autonomously. Key capabilities include project scaffolding via `quarkus_create`, loading extension-specific coding patterns via `quarkus_skills`, semantic documentation search, and proxying to the built-in Dev MCP tools. Extension authors can ship a `META-INF/quarkus-skill.md` file in their deployment JAR to teach agents the right patterns, pitfalls, and testing guidelines for their extension. A three-layer skill composition system (extension, user-level, project-level) allows teams to override or enhance default guidance. The server runs entirely locally with no telemetry.