Guide · installing Hermes Agent

Install Hermes Agent. The steps, and the security framework that goes with them.

Hermes Agent installs in a few minutes on Linux, macOS, Windows or WSL2. This guide covers the high-level steps and the decisions to make; the exact commands are in the official documentation, which changes with every release.

Hermes Agent is an open source project by Nous Research. Kailli is not affiliated with Nous Research.

01The steps

Install Hermes Agent in five steps

Order matters: you decide on the framework before granting access.

  1. Choose where it runs

    On a laptop to try it out, on a server for a team. Hermes Agent can run its commands locally, in Docker, over SSH or in a remote sandbox.

  2. Install from the official documentation

    Installer for Linux, macOS and WSL2, native Windows version, desktop app or Docker image: follow the official Hermes Agent documentation for the command for your system.

  3. Choose the model

    Nous Portal, OpenRouter, OpenAI, Anthropic, Google or a local model: the choice depends on your data, the quality you expect and your budget.

  4. Connect messaging apps and tools

    Discord, Slack, Telegram, email or Microsoft Teams to talk to it; MCP servers to connect it to your tools. One connection at a time, tested before the next.

  5. Set the rules

    Who can talk to it, what it can read, which commands need your approval. Write these rules down before the first real project.

02Skills and memory

Hermes Agent skills and memory

This is what sets it apart from an assistant: it learns know-how and remembers context.

Skills

A skill is a know-how card the agent loads when it needs it: publishing an article, preparing an invoice. Hermes Agent creates them from experience, improves them through use, and follows the open agentskills.io standard.

Memory

It keeps lasting notes about you and your environment, and can search its past sessions. Review what it retains, like a shared notebook.

Scheduled tasks

A built-in scheduler runs tasks at set times and delivers the result to your messaging app: Monday summary, news watch, backup checks.

03Security

Securing Hermes Agent in a business

An agent that acts needs written rules. Hermes Agent provides the mechanisms; it is up to you to configure them.

  • Human approval: dangerous commands wait for explicit approval; keep this mode on.
  • Authorized users: only listed people can talk to it in messaging apps.
  • Secrets: keys and passwords stored outside its working files, never pasted into a conversation.
  • Permissions: a dedicated account, read access first, write access only where it is useful.
  • Isolation: on a server, execution inside a Docker container rather than on the host machine.

The security mechanisms are described in the official Hermes Agent documentation, reviewed on October 17, 2026.

Guided installation: we install Hermes Agent with you, security framework included.

Book 20 minutes

Written by Rémi Carbonnefounder and CTO of Kailli

Frequently asked questions about installing Hermes Agent

How do you install Hermes Agent?

With the official installer or the desktop app, on Linux, macOS, WSL2 or Windows. The exact command for your system is in the official Hermes Agent documentation.

Does Hermes Agent run in Docker?

Yes. It can run in Docker and use Docker to execute its commands, which isolates its actions from the host machine. That is our choice on servers.

What is a skill in Hermes Agent?

A know-how card the agent loads when it needs it. It can create skills from experience and improve them through use. Skills follow the open agentskills.io standard.

Can Hermes Agent act without permission?

Commands deemed dangerous require explicit approval as long as approval mode is on. We recommend leaving it on and limiting the account's permissions.

How long does a guided installation take?

It depends on your tools and your infrastructure. We set the scope together during an initial 20-minute call.

Twenty minutes. And you'll know whether we can help.

No sales pitch. You tell us where you stand with AI, we tell you how far you can go.