🔗 Why MCP was always a bad idea

MCP was hailed as a necessary protocol when released by Anthropic in 2024, but the landscape has drastically changed. As LLMs became adept at executing code and autonomously calling APIs, MCP began to feel redundant. While it facilitated early integrations between AI models and external services, the evolution of these models means they can now interact directly with APIs without reliance on MCP, which often wraps existing services. The recent emergence of Cloudflare's Code Mode exemplifies how LLMs have outpaced the capabilities offered by MCP, composing scripts directly and omitting the need for intermediary servers. The article argues for a shift towards direct use of HTTP APIs and standard practices. Standardization could streamline interactions, making agents more efficient than ever before. Thus, embracing closure on MCP may very well be a strategic step forward in AI development.