Self-Evolving Hermes Agents: Enterprise AI That Gets Better With Use | Nemotron Labs

Most organizational knowledge lives informally — in people's heads, review comments, and repeated workflows. Hermes learns those patterns and turns them into skills that improve with every interaction. In this session, we're joined by Nous Research — the team behind Hermes Agent, a self-evolving AI agent built with the NemoClaw blueprint. See a live demo of Hermes in action, turning real work into reusable skills, inside a sandbox with policy you can read and trust. What you'll learn: How NVIDIA OpenShell gives Hermes a more secure foundation to operate in, with declarative policy you control How Hermes turns repeated work into skills: procedural knowledge the agent writes and maintains for itself How sessions, search, and curated memory work together to give Hermes a working knowledge of you and your projects Building something similar or curious about running your own agent harness inside NemoClaw? Drop your questions live — the Nous Research and NVIDIA teams are both here to answer them.