Hello legends — in this video I walk you through a full Hermes Agent setup from scratch, step by step. We start at the Hermes GitHub, run the install, and I show you the “full setup” path so you actually understand what each setting does. Big callout here: I don’t recommend running Hermes on your main computer. If you’ve got private files, passwords, banking logins, etc., put this on a separate device so if anything ever goes sideways, you’ve isolated the risk. From there I plug Hermes into a fully local model using Ollama (no paid APIs). I’ve been testing Gemma 4 a lot recently, and I explain why the smaller E2B model disappointed me for basic tool-following, while the E4B (9.6GB) handled the web search workflow properly. Then we wire up self-hosted Firecrawl for private web search (Docker required), plus Telegram so you can chat with your agent like a normal bot. I also cover the practical settings that actually matter in real use—tool iteration limits, showing tool-call progress, context compression, and session reset—because if you don’t manage context, these smaller local models get dumb fast. By the end, you’ve got a fully local, fully private Hermes agent doing web search and replying in Telegram.