Anthropic keeps advertising its MCP (Model Context Protocol), but to the extent it doesn't support other LLMs, e.g. GPT, it couldn't possibly gain adoption. I have yet to see any example of MCP that can be extended to use a random LLM.