Hermes Agent by Nous Research is a free, open-source AI agent that remembers what you tell it, writes its own skills, uses real tools and keeps working when you close the laptop.
Nous Research is the AI lab that builds it, and the code is on GitHub under the MIT licence, so the software costs nothing.
This is the operator walkthrough, so you'll get who the company is, what the agent does, what it costs and the exact commands to install it and set it up.
I checked every fact below on 12 October 2026 against the Nous Research website, the official Hermes docs, the GitHub API and my own install, and I'll tell you where each one comes from.
Hermes Agent By Nous Research At A Glance
| Question | Short answer | Where I checked |
|---|---|---|
| What is it? | It is a self-improving AI agent that runs on your machine, a server or the cloud. | The official README and docs |
| Who makes it? | Nous Research, Inc. makes it. | The Nous Research website |
| What licence is it? | It uses the MIT licence. | The GitHub API |
| What does it cost? | The software is free, and the model is the only part that can bill you. | The docs and Nous Portal |
| When did it launch? | The repository was created in July 2025 and the project went public in late February 2026. | The GitHub API and the v0.2.0 release notes |
| What is the latest release? | The latest tagged release is v0.21.6, published on 8 October 2026. | The GitHub API |
| How popular is it? | The repository showed about 252,000 stars and 54,000 forks. | The GitHub API on 12 October 2026 |
Who Nous Research Is
Nous Research describes itself on its own website as "a pioneer in open AI training and research".
Its stated mission is to "create and proliferate open access to intelligence".
The company's legal name in the website footer is Nous Research, Inc.
Its GitHub organisation was created in May 2023 and held 113 public repositories when I looked.
The press has covered the business side, so here is what reputable outlets reported.
- The Block reported in April 2025 that the company was founded in 2023.
- The same article named the founding team as Jeffrey Quesnelle as CEO, Karan Malhotra, a pseudonymous researcher called Teknium and Shivani Mitra.
- The Block and Fortune reported a $50 million Series A round led by the investment firm Paradigm in April 2025.
- The Block also reported about $20 million of earlier seed funding, including a $5.2 million round in January 2024.
- The same coverage described Psyche, which is a network Nous is building to train AI models across many separate computers.
I have left out anything about the company that I could not confirm in those sources.
The Hermes Models vs The Hermes Agent
This is the part that confuses most people, so I'll keep it simple.
Nous Research uses the name Hermes for two different things.
| Hermes models | Hermes Agent | |
|---|---|---|
| What it is | A family of open-weight language models | A program that runs an AI agent |
| Where it lives | Hugging Face, under the NousResearch account | GitHub, at NousResearch/hermes-agent |
| First appeared | Nous-Hermes-13b was uploaded in June 2023 | The repository dates from July 2025 and went public in early 2026 |
| Examples | Hermes 3, Hermes 4 and Hermes 4.3 | Version 0.21.6 in October 2026 |
| Do you need it? | You don't need a Hermes model to run the agent | You don't need the agent to run a Hermes model |
A model is the brain that writes the answers.
An agent is the body around the brain, which gives it memory, tools, a schedule and a way to reach you.
Hermes Agent doesn't lock you to a Hermes model, and the official README says you can use Nous Portal, OpenRouter, OpenAI or your own endpoint.
That means you can run Hermes Agent on a Claude model, a GPT model or a free local model, and many people do exactly that.
🔥 Want my exact Hermes Agent By Nous Research setup, step by step? Inside the AI Profit Boardroom you get step-by-step video trainings, my installable Agent OS and four coaching calls a week with 3,400+ members building real automations. → Join the AI Profit Boardroom
What Hermes Agent Actually Does
The README calls it "the self-improving AI agent built by Nous Research".
Here is every main part, in plain words, with the docs as my source.
- Memory means the agent keeps notes about you and your work between sessions, and it can search its own past conversations.
- Skills are saved instructions the agent writes for itself after a hard task, and it improves them as it uses them.
- Tools let it act, and the docs list more than 60 built-in tools for things like web search, browsing, vision, image generation and speech.
- The gateway connects one agent to more than 20 messaging platforms, including Telegram, Discord, Slack, WhatsApp, Signal, email and Microsoft Teams.
- Scheduled jobs let you describe a recurring task in plain English and have the result delivered to any of those platforms.
- Subagents let it hand parts of a job to isolated helpers that work in parallel.
- Terminal backends decide where commands run, and the docs list seven, which are local, Docker, SSH, Singularity, Modal, Daytona and Vercel Sandbox.
- Profiles are separate agents on one machine, each with its own model, memory, skills and keys.
- Plugins add extra powers, and the command line has a curated catalogue you can search and install from.
- Voice lets you talk to it and hear it answer as it thinks.
- Hermes Desktop is the official app with a normal window, so you never have to open a terminal.
- Hermes Cloud is the official hosted version, where Nous runs an always-on agent for you.
My sibling post on the Hermes Agent backend options goes deeper on where the commands run.
Step 1: Install Hermes Agent
You have two routes, and both are official.
The easy route is Hermes Desktop, which you download from the Hermes website.
The Nous Research homepage lists downloads for Mac, Windows and Linux.
The terminal route is one command on Linux, macOS or WSL2.
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
On native Windows, the README gives this PowerShell command instead.
iex (irm https://hermes-agent.nousresearch.com/install.ps1)
The docs also cover Nix and an Android package for Termux, which I haven't used myself.
After the install finishes, reload your shell and check that it worked.
hermes --version
hermes doctor
On my Mac, hermes --version printed a v0.21.4 build on the day I wrote this, which is slightly behind the latest tag because I update on my own schedule.
hermes doctor checks your configuration and tells you what is missing.
Step 2: Give Hermes Agent A Brain
Hermes Agent has no built-in model, so the first real choice is which brain to plug in.
- Run
hermes setupto open the interactive wizard. - Run
hermes setup --portalif you want to log in to Nous Portal and pick a model in one go, which is the first step the docs recommend. - Run
hermes modellater whenever you want to change the provider or the model. - Run
hermes fallback addto add a backup model, so a rate limit doesn't stop your work.
Nous Portal is Nous Research's own model service, and its pricing page showed four plans when I checked.
| Nous Portal plan | Monthly price | What the page says you get |
|---|---|---|
| Free | $0 | Free models only, with standard rate limits |
| Plus | $20 | $22 of monthly credits and access to 200+ models |
| Super | $100 | $110 of monthly credits |
| Ultra | $200 | $220 of monthly credits |
Free models rotate, so don't build your business on one of them.
When I recorded my latest video, typing "free" into the model list showed about eight free models, including a stealth model called MissingNo.
The Nous Research homepage described MissingNo as free for a limited time, and my Hermes MissingNo walkthrough covers how I tested it.
Step 3: Say Hello And Teach It About You
Type hermes and you're in a chat.
- Tell it who you are, what your business does and how you like answers written.
- Ask it to save that to memory.
- Give it one small real task, such as summarising a file in the folder you're in.
- Type
/skillsto see what it already knows how to do. - Type
/learnfollowed by a topic or a document when you want it to write a new skill.
I've used /learn to turn a whole book into a skill the agent can reuse, and it's the feature I show people first.
The first session always feels generic, and that's normal.
The agent gets more useful as its memory fills up, which is the whole point of the design.
🔥 Want me to walk you through this setup step by step? Inside the AI Profit Boardroom, there's a full Hermes Agent course plus four live coaching calls a week where you can share your screen and get your setup fixed. You'll be building alongside 3,400+ members. → Get access here
Step 4: Add Profiles, So One Machine Runs Many Agents
A profile is a separate Hermes home with its own config, keys, memory, sessions and skills.
hermes profile create coder
hermes -p coder setup
hermes -p coder
hermes profile list
I keep one profile per model inside my Agent OS, which means switching brains is one flag.
It also means I can compare a new free model against my usual one without touching my main setup.
Don't point two running agents at the same profile, because both will write to the same memory.
Step 5: Put Hermes Agent On Your Phone With The Gateway
The gateway is the piece that makes Hermes feel like a teammate.
- Run
hermes gateway setupand pick a platform such as Telegram. - Follow the prompts to paste the bot details that the platform gives you.
- Run
hermes gateway installto add it as a background service. - Run
hermes gateway start, then check it withhermes gateway status. - Run
hermes pairingto approve who is allowed to message your agent.
The security docs say gateway access is denied by default, which is what you want.
An unknown person who messages your bot gets a one-time pairing code, and nothing happens until you approve it.
Step 6: Schedule Work With Cron
Scheduled jobs are where the agent starts saving real time.
Run hermes cron create and describe the job in plain English.
These are three jobs I run every day.
- A 6am recap sends me my to-do list for the day.
- A daily Japanese lesson arrives without me asking.
- A morning inbox summary tells me which emails need a reply.
The command I'd memorise is hermes pause, because it's the emergency stop for cron jobs and new gateway turns.
hermes resume lifts the stop when you're ready.
🚀 Ready to put Hermes Agent By Nous Research to work in your business? Inside the AI Profit Boardroom you get step-by-step video trainings, my installable Agent OS and four coaching calls a week with 3,400+ members building real automations. → Join the AI Profit Boardroom
Step 7: Add Plugins, Skills And A Dashboard
Do this last, and add one thing at a time.
hermes plugins search voice
hermes plugins install <name>
hermes plugins list
hermes dashboard
hermes plugins browse lists every entry in the curated catalogue.
When I covered the plugin directory in Hermes Desktop, it showed around 500 free community plugins with ratings and a one-click install.
Community plugins are written by strangers, so read the GitHub page before you install one.
hermes dashboard opens a web dashboard on your own machine, and it listens on port 9119 by default.
My Hermes Agent plugins marketplace post shows the install flow with real examples.
What Hermes Agent Costs To Run
| Part | Cost | Notes |
|---|---|---|
| Hermes Agent | $0 | It is MIT-licensed open-source software. |
| Hermes Desktop | $0 | It is a free download. |
| Nous Portal Free plan | $0 | It covers free models only. |
| Nous Portal paid plans | $20, $100 or $200 a month | The plans include monthly credits. |
| Hermes Cloud Standard | $0.51 a day while running | The instance has 2 GB of RAM and 2 vCPUs. |
| Hermes Cloud Pro | $0.99 a day while running | The instance has 4 GB of RAM and 4 vCPUs. |
| Hermes Cloud stopped | $0.03 a day | You pay for storage only. |
| A local model | $0 | The price is paid in hardware. |
The Hermes Cloud page said you need at least $2 in credits, or an active subscription, to deploy a Cloud Agent.
Those cloud prices don't include model usage or tool usage, so the real bill depends on the brain you pick.
How Fast Hermes Agent Ships Updates
The release history on GitHub is public, so the pace is easy to measure.
The first tagged public release was v0.2.0 on 12 March 2026.
Its notes say the project had gone "from a small internal project to a full-featured AI agent platform" in just over two weeks.
By 8 October 2026 the project had shipped 37 tagged releases and reached v0.21.6.
Some releases get names.
| Release | Date | Name |
|---|---|---|
| v0.13.0 | 7 May 2026 | The Tenacity Release |
| v0.15.0 | 28 May 2026 | The Velocity Release |
| v0.16.0 | 5 June 2026 | The Surface Release |
| v0.18.0 | 1 July 2026 | The Judgment Release |
| v0.19.0 | 20 July 2026 | The Quicksilver Release |
| v0.20.0 | 3 August 2026 | The Herald Release |
The Herald Release notes list about 3,650 commits from more than 650 contributors, and it added streaming voice, signed webhooks, agent-to-agent messaging and cited research.
I tested that release inside my own setup, and my Hermes v0.20 guide has the results.
The practical lesson is to run hermes backup before hermes update, because a project that moves this fast will sometimes move your settings.
How To Keep Hermes Agent Safe
"Is Hermes Agent safe?" is the question I get most, and the honest answer is that it depends on your settings.
The official security page describes these protections.
- Risky commands go through an approval step, and the default mode uses a second model to judge the risk.
- A hard blocklist stops catastrophic commands, and no setting turns it off.
- Chat users are denied by default until you allow or pair them.
- The Docker backend runs commands in a locked-down container.
- Project files are scanned for prompt-injection patterns before they are loaded.
The same page is honest about the limits.
It says the file-write guards are defence in depth and not a hard boundary.
It also says turning approvals off should only happen in trusted environments.
My rules are simple.
- Keep approvals on.
- Give it read-only access first.
- Run
hermes securitynow and then, because it audits your plugins and tools against known vulnerabilities. - Never paste a secret into a chat that other people can read.
Hermes Agent Compared With Other Agents I Cover
| Agent | Type | Cost | Runs where |
|---|---|---|---|
| Hermes Agent | Open-source personal agent from Nous Research | Free, MIT | Your machine, a server or Hermes Cloud |
| OpenClaw | Open-source personal agent | Free, MIT | Your machine or server |
| Claude Code | Paid coding and work agent from Anthropic | A paid Claude plan or API | Your machine |
| OpenCode | Open-source coding agent | Free, MIT | Your machine |
| Hark Pro | Hosted personal agent from Hark Labs | A free plan, then $20 or $100 a month | Its own cloud computer |
| StarNet | Open-source agent station | Free, MIT | Your machine |
Hermes is the one I build on, because memory, skills, chat apps and scheduling all come in the box, and my whole Agent OS runs on it.
My best Hermes Agent alternative ranking covers the others one by one.
What Is Official And What Is Not
A big open-source project attracts add-ons, and some of them use the Hermes name.
| Product | Made by Nous Research? | What it is |
|---|---|---|
| Hermes Agent | Yes | The open-source agent |
| Hermes Desktop | Yes | The official desktop app |
| Hermes Cloud | Yes | The official hosted agent |
| Nous Portal | Yes | The official model service |
| Herald OS | No | An independent open-source desktop that uses Hermes as its interface |
| FlyHermes | No | A paid hosting service run by HF Data Studio LLC |
Herald OS and FlyHermes both state on their own pages that they are not affiliated with Nous Research.
The name overlap is easy to trip on, because version 0.20 of Hermes Agent was called the Herald Release.
Who Hermes Agent Is For
Hermes Agent suits you if you want one assistant that learns your business and works on a schedule.
It suits operators who are happy to type a few commands, or who will use the desktop app.
It suits agencies that want a separate agent per client, because profiles make that clean.
It's a weaker fit if you only want a coding tool inside your editor, because a dedicated coding agent is simpler for that.
It's also a weaker fit if you want zero setup and zero maintenance, because a fast-moving open-source project needs a little care.
🚀 Want the finished version instead of building it yourself? The AI Profit Boardroom includes my installable Agent OS, with Hermes profiles, the memory vault and the dashboard already wired together. You also get four coaching calls a week and 3,400+ members to help when you're stuck. → Join the AI Profit Boardroom
Related Reading
- The best Hermes Agent setup ranks six ways to put these pieces together.
- My Hermes MissingNo walkthrough covers the free stealth model on Nous Portal.
- The Hermes Agent backend guide explains where your agent's commands actually run.
Also On Our Network
- 🌐 What a business actually gets from Hermes Agent
- 🌐 Hermes Agent reviewed and scored, feature by feature
- 🌐 The seven-step Hermes Agent quick-start for an Agent OS
- 🌐 The models that run inside Hermes Agent, with real scores
FAQ: Hermes Agent By Nous Research
What is Hermes Agent by Nous Research?
Hermes Agent is a free, open-source AI agent built by Nous Research, an AI lab that describes itself as a pioneer in open AI training and research.
The agent keeps memory, writes its own skills, uses tools, runs scheduled jobs and talks to you through the terminal, a desktop app or chat apps such as Telegram.
Is Hermes Agent free and open source?
Yes, the GitHub API lists the NousResearch/hermes-agent repository under the MIT licence, and the install costs nothing.
The model you connect is the only part that can bill you, and Nous Portal has a free plan that is limited to free models.
Is Hermes Agent the same as the Hermes models?
No, because Hermes is also the name of Nous Research's family of open-weight language models, such as Hermes 3 and Hermes 4.
Hermes Agent is the software that runs an agent, and it works with almost any model, including ones that have nothing to do with Nous Research.
How do I install Hermes Agent?
Download Hermes Desktop from the Hermes website, or run curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash on Linux, macOS or WSL2.
On native Windows the official PowerShell command is iex (irm https://hermes-agent.nousresearch.com/install.ps1).
Then run hermes setup to pick a model provider.
Is Hermes Agent safe?
It is as safe as the access you give it.
The official security docs describe command approval, deny-by-default access for chat users, container isolation and scanning of context files.
The same docs say some guards are defence in depth and not a hard boundary, so start with approvals on and limited access.
Is Herald OS or FlyHermes made by Nous Research?
No, because Herald OS is an independent open-source desktop by a developer called Luke The Dev, and FlyHermes is a paid hosting service run by HF Data Studio LLC.
Both say they are not affiliated with Nous Research.
The official products are Hermes Agent, Hermes Desktop, Hermes Cloud and Nous Portal.
📺 Video notes + links to the tools 👉
🎥 Learn how I make these videos 👉
🆓 Get a FREE AI Course + Community + 1,000 AI Agents 👉
About Julian
I'm Julian Goldie, an SEO entrepreneur, author and founder of the AI Profit Boardroom, which has 3,400+ members.
I help business owners scale with AI agents, automation and SEO.
- I built Goldie Agency, a 7-figure SEO and link building agency based in Bangkok.
- I've grown a YouTube channel to 400,000+ subscribers.
- I wrote "Link Building Mastery", a book on link building strategies.
- I run GoldieBench, where I score AI models on real one-shot builds.
→ Get my best AI training inside the AI Profit Boardroom
Install it, give it a brain, teach it about your work, and you'll understand in one afternoon why so many operators build on Hermes Agent by Nous Research.
