If requests are already hitting your servers, it’s too late — you’re burning CPU and bandwidth before rejecting. I built Rately to solve this at the edge: set limits by user, IP, or plan (e.g. free = 1k/day, pro = 1M/month) and return 429s before traffic reaches your infra. Keeps apps simple and protects backend capacity.