
What Hermes Agent is
Released in February 2026, Hermes runs in your terminal, a native desktop app, your messaging platforms and IDEs. It's provider-agnostic — it works with 20+ model providers, including local models — and it sandboxes the actions it takes so it can work safely.
Its defining feature is the learning loop: when Hermes solves a problem well, it can save the approach as a skill and reuse it next time. Over weeks it builds up a library of how your business runs — which is a genuinely different experience from a stateless chatbot.
What it does well
Learning loop
Creates skills from experience and improves them during use — the more you use it, the more capable it becomes.
Persistent memory
Remembers preferences, projects and past solutions, with pluggable memory backends.
Multi-platform gateway
One agent, one memory, across Telegram, Discord, Slack, WhatsApp, Signal, Email and more.
Many surfaces
CLI, terminal UI, native desktop app, web dashboard and IDE integration.
Provider-agnostic
Swap models and providers mid-workflow; rotate credentials across multiple keys.
Isolated sandboxing
Five backends — local, Docker, SSH, Singularity, Modal — with container hardening.
How you get it
Install via a shell installer, then run the setup wizard to pick your model and provider:
# macOS / Linux
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
# Windows (PowerShell)
irm https://hermes-agent.nousresearch.com/install.ps1 | iex
# Interactive chat / setup wizard
hermes
hermes setupHermes also ships a native desktop app for macOS, Windows and Linux if you prefer a graphical interface over the terminal. As with OpenClaw, the practical path for many businesses is a managed setup — configured, sandboxed and connected for you, then simply used from your phone or desktop.
Security & control
Hermes is built with control in mind: sandboxed execution backends, secret redaction, approval modes, and profiles that keep different contexts (work, personal) isolated. Native local-model support means you can run the reasoning on your own hardware if your data-protection requirements demand it.