Split a gateway rate limiter with an agent

Your Go API gateway handles about 50,000 requests a second. You need a new limiting layer: a token bucket per API key stored in Redis, a fallback for when Redis is unreachable, limits per plan from config, Prometheus metrics and tests. In a token bucket, each key holds up to N tokens that refill at a steady rate, and each request spends 1.

A coding agent will do much of the work.

How do you split the job? Decide what the agent builds, what you keep, and where you stop it for review.