Evaluation criteria table for selecting Model Context Protocol (MCP) gateways and AI agent infrastructure. The table ranks key decision factors by research frequency and what to verify, including time to first working connection, OAuth and multi-user authentication complexity, self-hosting burden, policy and tool-level permissions, debugging and observability, production reliability, integration maintenance, pricing predictability, vendor lock-in and migration, and documentation and support. Each criterion highlights practical evaluation areas such as transport compatibility, tenant isolation, Kubernetes deployment, failover, token refresh, configuration portability, and version accuracy for enterprise AI deployments.