What would make devs choose an agent builder over LangGraph or n8n?
A developer building a visual AI agent builder with 1,600+ MCP tools asks what would actually make developers choose a platform over rolling their own with LangGraph or n8n. Tool count looks like a vanity metric; the thread argues the real decision comes down to debugging, evals, durable human approvals, and an escape hatch to code. The hard part, everyone agrees, starts after the demo works.