Hermes Agent: what it is, how to install it and what you maintain yourself

What Hermes Agent is

Hermes Agent is made by Nous Research, the team behind the open Hermes language models. The project describes itself as an agent that grows with you: it learns your projects, creates skills on its own and remembers how it solved a task. The code is open under the MIT license, and the software is free.

Hermes is a harness: the connected model does the thinking, and Hermes gives it tools, memory and a link to you. How such agents work is covered in How AI agents work.

• Any model

  Nous Portal, OpenAI, Anthropic, DeepSeek, OpenRouter or your own model via Ollama. A context window of 64,000 tokens or more is required.

• Memory

  Keeps conversations, notes and skills, so you can return to the work in the next session.

• Skills

  A skill is a SKILL.md instruction file. You install them from a catalog, and the agent creates new ones itself after solved tasks.

• 20+ platforms

  One gateway connects Telegram, WhatsApp, Discord, Slack, Signal, email and others.

• Scheduled tasks

  Reports, backups and summaries on a schedule set in plain words. They run without you, through the gateway.

What you need to install it

• A computer or server. Hermes runs on Linux, macOS, Windows, including via WSL2, and on Android via Termux. For the agent to answer around the clock, it goes on a server that never switches off; why a laptop is a poor fit for this is covered in Local AI agent.
• A terminal. Apart from the Hermes Desktop app, everything is done with commands: installation, model, gateway, checks.
• A model: a subscription login or an API key from one of the providers. Options are below.
• A Telegram bot or an account in another messenger to talk to the agent.
• Time for upkeep: updates, access, backups.

How to install Hermes Agent: five steps

1. Installation. On Linux, macOS and WSL2 run curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash; on Windows use the PowerShell command from the project page. There is also the Hermes Desktop app for Windows and macOS. After installing, restart the terminal.
2. Model. hermes setup offers three modes: quick via Nous Portal, full with your own keys, and a “blank slate” with almost everything off. To pick a provider step by step, use hermes model.
3. First conversation. hermes or hermes --tui opens a chat in the terminal. Check that the agent answers, uses tools and remembers the previous message. hermes -c continues the last session.
4. Telegram. Create a bot with @BotFather and save the token, find out your numeric ID, run hermes gateway setup, choose Telegram and enter the token and your ID in the allow list. hermes gateway starts the gateway.
5. Running without you. The gateway must run constantly, including after a server reboot. How to keep it in the background is described in the project docs; hermes gateway status shows its state.

Without an allow list the Telegram gateway lets no one in, and that is right. Keep the bot token secret: if it leaks, revoke it with /revoke at @BotFather. The full guide is in the Hermes Agent documentation.

Which models to connect to Hermes

Hermes works with a model whose context window is 64,000 tokens or more. There are three ways to connect one.

• Subscription login: Nous Portal, OpenAI Codex, Anthropic, xAI Grok and others.
• API key: OpenRouter, Fireworks, DeepSeek and other providers, billed per token.
• Your own model: Ollama, vLLM, SGLang or any server with an OpenAI-compatible API. What a model on your own hardware costs you is covered in Local AI agent.

Claude comes with a caveat. According to the Hermes docs, signing in via Anthropic works only on the Max plan and requires purchased extra usage credits: Hermes usage is billed to them, and the plan's own allowance is not used. A Pro subscription cannot use this path; without Max the project recommends an API key billed per token. This is Hermes's own rule. As of October 28, 2026, the Anthropic help article says that third-party apps signed in with a Claude plan draw from the subscription limits; in spring, from April 4, third-party harnesses required extra payment or an API key. A Claude Pro or Max subscription covers Anthropic's own products, including Claude Code, within the plan's limits.

What you will maintain yourself

With an agent you host yourself, upkeep begins after installation. With Hermes it comes down to five jobs.

• Updates. Install new versions and check that the agent, skills and gateway work after the update.
• Access. The bot token, model keys in ~/.hermes/.env, the allow list. If a key leaks, you replace it yourself.
• Server. Restarts after failures, disk space, network ports: nothing extra should face the internet.
• Backups. Memory and skills build up for months, and you have to back them up yourself.
• Model costs. With per-token billing the bill depends on the volume of tasks, and it needs watching.

This is manageable if you already administer a server. If not, set aside time: the first weeks go into getting the agent to run reliably.

Who Hermes Agent suits

• You are confident in the terminal and want to build an agent your way

  →Hermes Agent on your own server

• You want to try an agent on your computer in an evening

  →Hermes Desktop on Windows or macOS

• You need an everyday assistant and have no time to fix it

  →A turnkey agent

Hermes is a strong open-source project, and the difference between the paths is the same as with other open-source agents: who installs, fixes and updates it. Types of agents are compared in Which AI agent is best, and what can be free in an agent in Free AI agents. In NekoAgent we do the first install together on a call with a curator, and then during paid support Iris monitors the agent around the clock and helps bring it back after failures (reasonable efforts, no guaranteed response time), makes nightly encrypted backups and ships tested updates you install with one button.

An agent without the terminal

Plans from 24,900 RUB for three installs. The first one is on a call with a curator, with training.

Plans and prices Ask a question

Key questions