Most MCP servers don't need to exist. Your case might be an exception
The article argues that most MCP (Model Context Protocol) servers are unnecessary, as many use cases can be handled more simply with existing tools or direct API calls. However, it acknowledges that complex, multi-step workflows requiring external data or actions may still justify building a dedicated MCP server.
Background
- MCP (Model Context Protocol) is an open standard developed by Anthropic that lets AI assistants connect to external tools, data sources, and services through "MCP servers" — basically middleware that translates between an LLM and an API or database.
- The article argues that in most cases, you don't need a dedicated MCP server. The same result (an AI tool-calling an API) can be achieved more simply by giving the LLM a well-written function definition or using a generic HTTP tool, without the overhead of running and maintaining a separate server process.
- The exception: when you need to share a capability across multiple AI applications, enforce access controls, or handle complex, stateful workflows that require a persistent server. If you're just hooking up one chatbot to one API, an MCP server is likely overengineering.
- Evil Martians is a well-known engineering consultancy that publishes technical deep-dives on modern web development, infrastructure, and AI tooling. Their audience tends to be senior engineers evaluating new tools critically rather than following hype.