About a year ago, frameworks like CrewAI, Phidata, and LangGraph were everywhere. Now I barely hear about them, or really any “agent framework” at all. I’ve been trying to build my own AI agent and looked into OpenClaw it almost feels like its own framework. But it doesn’t seem like people are standardizing around anything. Are people actually using a common library right now? Or is everyone just rolling their own setups like custom wrappers around MCPs(more CLI now) , agent handoffs?, and things like skills.md? Would like to know what people are actually using in real projects.