LangChain, CrewAI… but which frameworks actually scale?

I’ve been trying out a bunch of agentic AI frameworks lately while hacking on projects. (Funny enough, I’m testing one right now as I write this post 🤓)

The ones I hear mentioned most are LangChain and CrewAI. Both are great for demos- but once I push into production use, I start seeing issues:

• Workflows crash halfway through

• Context/memory loss in multi-step tasks

• No real concurrency, everything runs like it’s single-threaded

• Debugging feels like untangling spaghetti

I’m curious, for those of you building agentic workflows:

👉 Which framework are you using today, and why?

👉 What’s been your biggest technical blocker moving from demo → production?

(For me, these pain points are exactly why we’re building GraphBit- a developer-first, enterprise-grade ultra efficient framework launching soon)