MCP Code Mode and Erlang: Building a True Google Workspace Universal "MCP" (sean.lyn.ch)

Cloudflare has introduced a novel approach to its MCP (Multi-Context Programming) tool with the launch of its Code Mode MCP, designed to enhance developer interactions with its expansive API of over 2,500 endpoints. Rather than exposing all endpoints, which could overwhelm users, Cloudflare leverages the strengths of large language models (LLMs) to generate code using a typed SDK. It employs two primary functions, `search()` to navigate the full OpenAPI spec and retrieve relevant endpoints, and `execute()` which runs JavaScript code in a secure, isolated environment. This innovative setup allows developers to dynamically create and execute code against Cloudflare's API while avoiding the usual complexities involved in setting up API interactions. The significance of this development extends to Google's ecosystem, where a similar universal server can be created using Google Apps Script. Unlike traditional methods that require cumbersome authentication setups and API configurations, this server operates by simply deploying a one-time script that can execute JavaScript on demand. With built-in OAuth token handling and automatic capability discovery via Google's Discovery API, the system allows users to access all Google Workspace functionality seamlessly. This reflects a broader trend in AI and cloud services: creating flexible, resource-efficient solutions that minimize deployment overhead by enabling real-time code execution and adaptability in workflows.