mcp.hosting: Simplify MCP server management with cloud-synced configuration

The Model Context Protocol (MCP) is a powerful standard, but managing it is currently a fragmented chore. Every time you add a tool, you're hunting for hidden JSON config files across different IDEs and clients. mcp.hosting addresses this specifically by acting as a configuration layer. By installing a small proxy (`mcph`), users shift the source of truth from local files to a cloud dashboard, allowing for instant synchronization across every connected client. It's a pragmatic solution to a real 'developer experience' friction point. The most compelling technical feature is the 'smart routing.' In a standard MCP setup, loading too many servers can pollute the LLM's context window and inflate token costs. By ranking and activating servers based on task requirements, mcp.hosting optimizes the prompt overhead. The decision to keep execution local—where `mcph` communicates directly with servers on the user's machine—is a critical security choice that prevents the cloud platform from becoming a bottleneck or a massive privacy risk for local data. However, the reliance on a third-party proxy introduces a new point of failure. If the mcp.hosting service goes down or the token expires, your toolset vanishes across all clients simultaneously. While the 'Team' tier offers self-hosting, the average user is trading local autonomy for convenience. The pricing is fair, though the 3-server limit on the free tier is a tight leash for developers who like to experiment with various community MCPs. This is a must-have for power users who jump between Cursor, Claude Desktop, and VS Code and are tired of copy-pasting environment variables. It transforms MCP from a manual configuration task into a manageable service.