Skip to main content
Before we wire an agent into this gateway, it’s useful to compare two approaches: building directly against the Kubernetes API with a custom script, or adopting a standardized protocol such as MCP. Both can work technically, but they come with different trade-offs around maintenance, interoperability, and security. MCP adoption has accelerated quickly. SDK downloads climbed from roughly 100k/month at launch in late 2024 to tens of millions/month by early 2026. The project is now under the Linux Foundation’s governance, and major vendors — OpenAI, Google, Microsoft, and AWS — are shipping support. That momentum makes MCP a compelling choice if you want a single, vendor-backed protocol to integrate with multiple tools without building and maintaining a separate integration for each one. That said, MCP is still maturing and has practical limitations today. The most significant is authentication: only a subset of MCP server implementations fully implement the auth specification. If you operate in a regulated environment or need strong, auditable authentication controls, evaluate the maturity of the MCP implementations you plan to use before relying on them for production workloads.
MCP provides standardization benefits (one protocol instead of many bespoke integrations). Verify the maturity of critical features—especially authentication—against your compliance and security requirements before committing.
Below is a quick comparison to help decide which approach fits your needs:
A man sits at a table speaking into a Rode microphone on a boom arm in a warmly lit podcast/studio setup with plants and a lamp behind him. He gestures with his hands as he talks.
Even while the ecosystem matures, MCP represents a sensible path forward for building AI agents that need to integrate with multiple cloud and vendor toolchains. If you choose MCP, plan for implementation validation (especially auth) and a migration strategy that allows swapping in more mature implementations as they become available.

Watch Video