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!
Replies
MindPal
We construct the function-calling schema from the API, and I have to say — it’s a daunting and error-prone task, especially since most of our customers don’t really understand these technical details. MCP simplifies this by allowing us to simply add its SSE URL and have everything set up automatically.
However, a new challenge has emerged: finding a reliable, dedicated remote MCP service to connect tools like Notion and Google Sheets. We’ve looked at options like Zapier and Make MCP, but I still feel like their MCP servers are built on top of their main services, not truly designed for agent-based architectures.