Paper accepted at ACM SIGCOMM 2025 NGNO Workshop

As AI agents become more capable of interacting with network environments, we see a growing need for standardized, reproducible, and open platforms that support experimentation and benchmarking.

In this work, we present a modular, extensible playground built on top of existing network emulators like Kathará, designed to simplify the evaluation of AI agents across diverse network scenarios.

Our goal is to call attention to the lack of holistic platforms in this space and to spark discussion around the challenges of benchmarking AI-driven solutions in networking.