The AI agent ecosystem is evolving: what were once simple API wrappers are now becoming systems capable of reasoning, learning, and taking action. In this context, how can we maintain control over the technology and our data? The answer lies in open source.

In this technical session, we will explore two opposite architectural philosophies through two frameworks that are gaining significant traction: OpenClaw and Hermes.

OpenClaw (the orchestrator): we will analyze the deterministic “CLI-first” approach. We will see how to manage multi-channel gateways and explicit skills via human-readable configuration files, ensuring that the agent performs exactly what it was designed to do.

Hermes (the learner): we will explore the autonomy aspect. We will see how Hermes uses self-learning cycles to generate new skills on-the-fly and adapt to user preferences, moving beyond the concept of static programming.

Finally, we’ll look at how to deploy applications at home. I’ll provide concrete tips on setting up a home lab for AI agents: from using quantized models to choosing hardware (RTX GPUs vs. Mac Silicon), all the way to managing local data persistence.

If you're a developer who believes in open source and want to learn how to scale your agents from your laptop to your home server, this talk will give you the tips you need to do so.