Hermes Agent is an open source AI agent you self-host — a persistent assistant from Nous Research that remembers what it learns across sessions, builds its own skills, and works with any LLM you point it at.

What is Hermes Agent?

Hermes Agent is a self-hosted AI agent from Nous Research that runs continuously on your own server. You message it from chat apps or a terminal, and it acts on your behalf using built-in tools. Its defining trait is a learning loop: it keeps memory across sessions and writes reusable skills from tasks it completes.

What is Hermes Agent best for?

People who want a private, always-on assistant they control end to end — one that lives on their own machine, remembers context, and grows more useful over time. It suits builders and operators comfortable bringing their own LLM keys and doing light server setup, rather than anyone wanting a zero-config hosted chatbot.

What can Hermes Agent do?

Persist across sessions: agent-curated memory plus full-text session search, so it recalls past conversations and a model of who you are
Build its own skills: after complex tasks it creates reusable skills and refines them during later use
Run 40+ built-in tools, including web search, browser automation, vision, image generation, and text-to-speech
Reach you anywhere: talk to it from Telegram, Discord, Slack, WhatsApp, Signal, Email, or the CLI through one gateway
Schedule recurring work in plain language and delegate parallel jobs to isolated subagents
Stay LLM-agnostic: connect Nous Portal, OpenRouter, OpenAI, Anthropic, or a custom endpoint and switch models with no code changes
Install in one command on Linux, macOS, or WSL2 and run on a small VPS (~2 GB RAM)

Is Hermes Agent free?

Yes — the agent itself is free and MIT-licensed, so you can self-host it at no cost. But it ships no model of its own: you connect an LLM provider (Nous Portal, OpenRouter, OpenAI, and others) and pay that provider per use, or run a local model on your own GPU. The software is free; the intelligence behind it is not.

Where does Hermes Agent fall short?

It isn’t a local-only solution. Hermes needs an external LLM provider and ongoing per-call API spend, unless you run your own model — which requires a GPU and more hardware than the lightweight agent itself.
It’s early-stage software. The project is pre-1.0 and under active development, with native Windows support still in beta, so expect rough edges and breaking changes between versions.
It executes code with real tool and terminal access on your server, so you have to handle sandboxing and security yourself before exposing it — more involved than a managed cloud assistant.

FAQ

Is Hermes Agent open source? Yes — it’s released under the MIT License by Nous Research, so you’re free to self-host, modify, and redistribute it.

Do I need my own LLM API key? Yes. Hermes is LLM-agnostic and brings no model of its own. You point it at a provider like Nous Portal, OpenRouter, or OpenAI (or a local model) and pay for that usage.

What do I need to run Hermes Agent? A Linux, macOS, or WSL2 machine — a small VPS with about 2 GB of RAM is enough — plus Python 3.11+, Node.js, and an LLM provider key. A single curl command installs the rest.

How is it different from a normal chatbot? It persists. Hermes runs continuously on your server, remembers across sessions, and writes its own reusable skills from tasks it finishes, so it gets more capable the longer you use it.