MCP Servers facilitate communication between AI clients and APIs like Semaphore's. Building an MCP Server requires minimal coding. The article walks through creating one that interacts with Semaphore's public API, which provides access to continuous integration and deployment (CI/CD) workflows. The setup involves creating a Python project and establishing a connection to Semaphore’s API using a service account for authentication. A simple curl command verifies that the API connection is functional before diving into server implementation. The server, primarily coded in a single file called `main.py`, uses the FastMCP library to define tools that the AI can call. One of these tools, `list_projects`, fetches project names from Semaphore. The function makes an authenticated GET request to the API and returns a list of project names in a structured format. Testing the MCP Server employs the MCP Inspector, an interactive tool that allows users to invoke functions and inspect results. Once the server runs successfully, it can be integrated with OpenAI's Codex, enabling conversational queries about Semaphore projects. Future developments include an official MCP server for Semaphore, which is in the works and will initially offer read-only access. This new server aims to enhance user interactions with Semaphore, allowing for more complex requests such as retrieving build logs or modifying deployment pipelines. The article illustrates the simplicity of connecting AI capabilities with CI/CD systems, making it easier for developers to manage their workflows interactively.