> ## Documentation Index > Fetch the complete documentation index at: https://notes.kodekloud.com/llms.txt > Use this file to discover all available pages before exploring further. # Why MCP and What Were Not Using Instead > Comparing adopting MCP versus custom Kubernetes API scripts for integrating AI agents, highlighting interoperability, maintenance trade offs, and authentication maturity considerations. 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](https://learn.kodekloud.com/user/courses/mcp-for-beginners). 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](https://learn.kodekloud.com/user/courses/mcp-for-beginners) 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](https://learn.kodekloud.com/user/courses/mcp-for-beginners) 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: | Consideration | MCP (standard protocol) | Custom script directly to Kubernetes API | | - | -: | - | | Interoperability | High — single protocol across vendors and tools | Low — per-tool integrations required | | Maintenance burden | Lower once ecosystem support exists | Higher — you maintain scripts for each tool | | Ecosystem momentum | Strong; growing vendor support and SDK downloads | Dependent on community or your team | | Authentication maturity | Varies by implementation — verify support | Fully under your control (but you must implement it) | | Best fit | Organizations wanting vendor neutrality and long-term scale | Environments needing immediate, tightly controlled auth or bespoke behavior | 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. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.