Open-source agent runtime by Nous Research

The agent that
grows with you.

Hermes is an autonomous AI agent that gets more capable the longer it runs. Work with it from your CLI, from Telegram, Discord or Slack — while it runs on a machine you never SSH into yourself.

Install Hermes Read the docs
  • 60+built-in tools
  • 20+messaging platforms
  • 7terminal backends
  • MITlicensed, on GitHub
Scroll

What is Hermes

Not a coding copilot tethered to an IDE. Not a chatbot wrapper around a single API.

It is an autonomous agent that gets more capable the longer it runs. It lives wherever you put it — a five-dollar VPS, a GPU cluster, or serverless infrastructure that costs nearly nothing while it sits idle.

Hermes creates skills from experience and improves them while it uses them. It nags itself to persist knowledge, keeps a deepening model of who you are across sessions, and recalls any earlier conversation through FTS5 search with LLM summarisation. Skills follow an open standard, so they stay portable and shareable with agentskills.io.

It is built by the model trainers at Nous Research — the lab behind the Hermes, Nomos and Psyche models — and works with Nous Portal, OpenRouter, OpenAI or any endpoint you already pay for.

Seven terminal backends mean the agent follows your infrastructure, not the other way around.

  • local
  • Docker
  • SSH
  • Daytona
  • Singularity
  • Modal
  • Vercel Sandbox

  • Memory system

    Persistent memory that grows across sessions, with cross-session recall.

  • Skills system

    Procedural memory the agent writes for itself — open standard, portable, shareable.

  • MCP integration

    Connect any MCP server, filter its tools, and extend the agent safely.

  • Voice mode

    Real-time voice interaction in the CLI, Telegram, Discord and elsewhere.

  • Security

    Command approval, authorization rules and container isolation.

  • Research-ready

    Batch processing, trajectory export and RL training with Atropos.

Features

A complete agent runtime, not a prompt wrapper.

Closed learning loop

Creates skills from experience, improves them during use, and nudges itself to persist what it learns.

Runs anywhere

Local, Docker, SSH, Daytona, Singularity, Modal or Vercel Sandbox. Serverless backends hibernate when idle.

Lives where you do

CLI, Telegram, Discord, Slack, WhatsApp, Signal and more — twenty-plus platforms from one gateway.

Bot Mode

A durable team of specialist bots with their own model, memory, skills and routines — they work together in group chats and through @mentions.

Scheduled automations

A built-in cron delivers results to any connected platform, on the schedule you set.

Delegates and parallelises

Spawn isolated subagents for parallel workstreams. Programmatic tool calling collapses multi-step pipelines into single inference calls.

Platforms

One gateway. Everywhere you already are.

Reach the same agent from whichever chat window is already open. Nothing to install on the client side.

  • CLI
  • Telegram
  • Discord
  • Slack
  • WhatsApp
  • Signal
  • Matrix
  • Mattermost
  • Email
  • SMS
  • DingTalk
  • Feishu
  • WeCom
  • Weixin
  • QQ Bot
  • Yuanbao
  • BlueBubbles
  • Home Assistant
  • Microsoft Teams
  • Google Chat

Getting started

Running in about a minute.

  1. 01

    Install the CLI

    One command on Linux, macOS, WSL2, native Windows, NixOS or Android through Termux. Prefer a GUI? Grab the desktop installer from the website.

  2. 02

    Connect a model

    hermes setup --portal walks you through one OAuth that covers a model plus web search, image generation, text-to-speech and the browser tool.

  3. 03

    Start talking

    Run it in the terminal, or wire up Telegram, Discord, Slack, WhatsApp, Signal and the rest from a single messaging gateway.

install 
$ curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash$ hermes setup --portal$ hermes

Free, open source and MIT licensed. Source and releases live on GitHub.