Full-Stack Optimizations for Agentic Inference with NVIDIA Dynamo | NVIDIA Technical Blog
…Agent framework that drives the workflow (Claude Code, Codex, OpenClaw, OpenCode, etc.) Orchestrator : Dynamo’s routing, scheduling, and cache management layer Runtime : Inference engine that executes the model and owns the KV…