How do you manage tool integrations when building LLM agents?

Hey devs 👋

I'm currently working on an LLM agent project — integrating with Gmail, Slack, databases...
Every new API integration feels like building everything from scratch 😵‍💫

I recently explored something called MCPHubs, which standardizes how agents connect to APIs, files, and databases. It's like cutting down tons of repetitive work.

Now I'm curious—

💬 How do you usually handle tool integrations for your agents?
Custom wrappers? LangChain? Function Calling? Other frameworks?
Any major pain points you've hit?

👇 Would love to hear your experiences and lessons learned!