A developer demonstrates building a Model Context Protocol server using Common Lisp and the Gemini CLI. The project uses the 40ants/mcp library to create a stdio-based server, showing that MCP's language-agnostic design allows AI tools to interface with non-Python codebases.

The setup involves Steel Bank Common Lisp (SBCL) and the Quicklisp package manager. A minimal 'Hello World' server was built and validated, connecting the local Lisp process to Gemini CLI as an MCP client. This proves the protocol works for legacy or specialized languages.

The author then used Gemini CLI to extend the server with new tools, including a symbolic differentiation function. This showcases a practical workflow where an AI assistant helps modify and test code in a non-mainstream language, broadening MCP's applicability beyond Python-centric AI development.