ScyllaDB MCP Server: What AI-Native Developer Distribution Looks Like

Perhaps I'm misunderstanding something, but this article appears to be misleading.

The title and content suggest that it's about AI-native distribution, and that MCP servers are the answer (i.e., the old world prioritized good docs; now, if you want to be discoverable by agents, you need an MCP server). But it assumes agents can just "discover" arbitrary MCP servers (as if Claude would "just know" that you have an MCP server).. which is not the case?