Are AI agent tools (like MCP servers) too fragmented right now?

Posted by DrawingFluffy9866@reddit | LocalLLaMA | View on Reddit | 7 comments

I’ve been trying to use MCP servers for local AI agents and honestly, discovery + setup feels messy.

For example:

- Found 5+ tools on GitHub → no clear docs or install steps

- Some don’t work with my setup (llama.cpp)

- No way to quickly test before integrating

Curious:

- Where are you actually finding reliable MCP tools?

- Do you just stick to a few trusted ones?

Feels like there’s a gap for something like a “verified MCP registry” with easy testing.

Am I overthinking this or are others facing it too?