The Self-Improving OSS Agent Stack — Marc Klingen, Langfuse
An OSS reference stack for self-improving agents combines online monitoring with offline evaluation cycles.
“I mean the transition from "I don't write prompts" to "I have cycles."”
LangFuse co-founder Marc Klingen describes an emerging OSS reference stack where teams close the loop between production agent monitoring and offline dataset evaluation. He argues that combining online observability with offline experimentation is now table stakes for serious agent development, replacing the old split between DataDog-style monitoring and MLFlow-style experiment tracking. The talk signals that agent quality engineering is maturing into a distinct discipline with its own toolchain.