Your AI agent is a distributed system
Originally aired on October 23 @ 10:00 PM - 10:30 PM EDT
AI agents look simple in a demo: give a model some tools and let it run. Production changes the problem completely. Agents need durable state, retries, scheduling, concurrency, real-time communication, observability, and a way to survive failures without losing their place. In other words, they start looking a lot like distributed systems. This session explores the infrastructure patterns behind reliable, long-running agents, what we’ve learned building them at Cloudflare, and how those lessons change the way developers should think about agent architecture.
English
