I developed a local MCP server in Rust that has some useful tool calls for often-needed commands. For example it can pull and compile a program, or load a blog post. It was interesting developing the MCP server, but I wanted to think about why an MCP server is superior to running the command lines directly.
The LLM (and the chat window) is a new interface for computer use. With a prompt I can chain multiple function calls together, instead of having to enter the commands one after the other. And prompts have some level of error correction built-in: a typo can be automatically corrected.
The agentic layer could also be used to allow voice commands to run the tools. So to restate:
• Using an AI agent with an MCP server to run commonly-needed commands may increase ease of use.
• Some development work is needed to get the MCP server working.
Currently I am testing Gemma 4 E4B to handle the agentic tool calls. I can control a lot of things I do on the computer through a chat window in llama-cpp
. I feel error correction, command chaining, and an improved UI (a browser instead of a terminal) are worthwhile.