Hermes Agent is MIT-licensed, open-source, and — per its own project site — designed with "no telemetry, no cloud lock-in." It runs on your own infrastructure and keeps a persistent memory of you across sessions instead of resetting after every conversation, which is the feature that separates it from both a plain chatbot and, per its own framing, from OpenClaw itself.
This is a recorded walkthrough, not a live terminal — showing the shape of a real install, not a substitute for the project's own current instructions.
recorded session — not a live shell0 / 3
curl -fsSL https://hermes-agent.org/install.sh | shOne command downloads and installs the agent — no separate dependency install step, per the project's own setup docs.hermes-agent initThis is where the model choice from lesson 2 actually gets made — Hermes Agent itself doesn't include a model, it calls whichever one you configure here.hermes-agent startOnce it reports "ready," it's live on whatever platform you connected — and from here on, every message is billed to the model backend you configured, exactly like OpenClaw.
A single command handles setup on Linux, macOS, or WSL2 — no prerequisites beyond having a shell.
Knowledge check
What does Hermes Agent add on top of what OpenClaw already does?