Use Cursor to deeply experience 3 MCP Servers. Is it amazing but not practical?

Deep Dive into 3 MCP Servers with Cursor: Impressive but Not Practical?

via Just For Fun

Telegraph
Deep Dive into 3 MCP Servers with Cursor: Impressive but Not Practical?

When large language models first emerged, they only generated response content through pre-trained models. At this time, the models had two main drawbacks: there was some data they didn't know, for example, a model trained in March 2024 wouldn't know about events in May 2024; and they couldn't use external tools. Here, tools can be equivalently understood as function calls. For instance, if I have a tool function for publishing articles, I can't use natural language to ask the large model to help me call this function. To solve these two problems, OpenAI was the first to support the function calling feature in the model. They introduced it in this blog: Function calling and other API updates…