silver cloak
Hey everyone,
With the recent shifts toward more agentic workflows, I know many of us are looking for ways to make tool-calling more standardized across different models. While OpenAI has a robust function-calling ecosystem, the Model Context Protocol (MCP) is becoming a great way to build tools once and use them everywhere.
If you’re working with the OpenAI Agents SDK or building custom assistants that need to pull from external data sources (GitHub, Google Drive, local DBs, etc.), I wanted to share CuratedMCP as a developer resource.
Why it’s useful for OpenAI builds:
Discovery: It’s essentially a directory of production-ready MCP servers, so you don't have to build every integration from scratch.
Managed Hosting: One of the hurdles with MCP is often the "local-only" nature of many servers. CuratedMCP provides the managed infrastructure (Gateways) to host these, making them accessible to your OpenAI API calls in a cloud environment.
Unified Integration: Since OpenAI now supports MCP through its SDK and Hosted MCP tools, you can use these curated servers to give your GPT-4o or GPT-5 agents specialized capabilities without managing the underlying server instances.
UI & Context: It aligns well with the new Skybridge/Apps SDK patterns for rendering rich tool outputs directly within agent interfaces.
I’m really curious to see how people are bridging the gap between standard OpenAI function calling and the broader MCP ecosystem. If you’re hitting any walls with hosting your toolkits or finding specific integrations, I’d love to chat or see if we can add what you need to the directory!
Happy building!
marsh bear