I most enjoy understanding the reasons behind how things work. A project using third-party MCP servers gave me a practical way to explore how tool calling connects a model to data held elsewhere.
I enjoy learning the ‘why’ behind how things work.
I worked on a project using third-party MCP servers.
The project explored how a database of 10,000+ restaurants could answer a question through tool calling.
I learned that the model never holds all that data.
In the project, the model decides which tool to call and with what parameters (like cuisine, location, or rating).
The server runs the query on its own side, and only a small, relevant result comes back.
I learned that good AI products depend as much on how the tools and data are structured as on the model itself.