Hermes Agent Installation Guide 2026 (Full Setup Walkthrough)

This hermes agent installation guide 2026 is the exact end-to-end setup I run when I drop Hermes onto a fresh machine, and it'll get you from zero to a working agent in under fifteen minutes.

I'll walk you through every install command, every post-install step, and every gotcha I've hit so you don't waste an afternoon debugging shell paths.

🔥 Want my full Hermes installation + workflow training? AI Profit Boardroom has the full Hermes setup course, Workspace + Swarm + Goal trainings, plus weekly live coaching. → Get the trainings

What Hermes Agent Actually Is

Hermes is the self-improving AI agent built by Nous Research and it's currently sitting on 145,000 GitHub stars for a reason.

It runs as a real terminal interface with a full TUI, multiline editing, slash-command autocomplete, conversation history, interrupt-and-redirect, and streaming tool output.

It lives where you do — Telegram, Discord, Slack, WhatsApp, Signal, Email, and the CLI all run off a single gateway.

It has a closed learning loop that creates skills from experience, improves them during use, searches its own past conversations, and builds a deepening model of who you are across sessions.

It runs anywhere — local, Docker, SSH, Singularity, Modal, Daytona, Vercel Sandbox — and you can host it on a five-dollar VPS or a serverless backend that costs nearly nothing when idle.

That's why this hermes agent installation guide 2026 matters so much — once it's installed, the rest of the AI agent landscape feels limited by comparison.

Install Command For Linux, macOS, And WSL2

Here's the one-liner that installs Hermes on Linux, macOS, and Windows Subsystem for Linux 2.

curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash

This single command pulls the official install script directly from the NousResearch GitHub repo and runs it through bash.

It installs every dependency Hermes needs including uv, Python 3.11, Node.js, and the supporting CLI tooling.

It writes the Hermes binary to your path so you can call hermes from anywhere on your machine.

If you're on a stock Ubuntu, Debian, or macOS install, this is the only line you need to remember.

Install Command For Windows PowerShell

Windows users get their own one-liner via PowerShell.

irm https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.ps1 | iex

This is the Windows-native installer and it currently ships as early beta — fully usable but you should know it's the newest of the three install paths.

The Windows installer handles uv, Python 3.11, Node.js, ripgrep, ffmpeg, and a portable Git Bash via MinGit.

MinGit unpacks to %LOCALAPPDATA%\hermes\git with no admin permissions required and stays completely isolated from any system Git you already have installed.

If you already have Git installed, the Hermes installer detects it and uses your existing install rather than downloading the 45MB MinGit bundle.

Either way, the installer won't interfere with your existing Git setup so it's safe to run on a developer machine.

Install Command For Termux On Android

Hermes runs on Android phones via Termux and the install is almost the same.

The catch is you need to use the curated .[termux] extra rather than the full .[all] because the standard package pulls in voice dependencies that aren't compatible with Android.

This is genuinely useful if you want a pocket-sized agent that runs on your phone and syncs with your other Hermes installs through the gateway.

I keep a Termux install for travel weeks where I don't want to lug a laptop but still need to dispatch agent work.

Post-Install Steps That Actually Matter

After the installer finishes, you need to reload your shell so the hermes command resolves.

source ~/.bashrc
hermes

If you're on zsh on macOS, swap that first line for source ~/.zshrc instead.

The moment you run hermes for the first time, the interactive CLI fires up and prompts you to pick a model and personality.

This is where most people stall — pick anything for now, you can change it later with one command, and you should never let the model picker block you from actually using the tool.

How To Choose A Model In Hermes

Hermes supports more than 200 models out of the box across Nous Portal, OpenRouter, NVIDIA NIM (Nemotron), Xiaomi MiMo, z.ai/GLM, Kimi/Moonshot, MiniMax, Hugging Face, OpenAI, and custom endpoints.

You switch models with one command.

hermes model

That command opens the model picker and lets you swap providers without touching any config files or restarting anything.

For most agent work I default to a Claude or GPT-class model — see my Sonnet 4.8 review for why Sonnet is my reasoning default in 2026.

If you want to run local models, install LM Studio or Ollama, run hermes setup, add the local provider, and then /model to switch.

The local route is what I recommend for anyone running Atomic Chat vs Ollama workflows where data privacy matters.

The Key Slash Commands You'll Actually Use

Hermes ships with a full set of slash commands that work in both the CLI and any messaging gateway.

/new or /reset starts a fresh conversation when context gets messy or you switch tasks.

/model [provider:model] swaps the underlying LLM mid-conversation without restarting.

/personality [name] swaps the agent's persona — useful when you want a more terse or more verbose response style for a specific task.

/retry and /undo let you redo or roll back the last turn, which is the equivalent of git undo for agent conversations.

/compress shrinks the conversation context when it's getting long, and /usage shows you exactly how much you've spent on the current session.

/insights --days N pulls back the agent's reflection on patterns from the last N days of conversations.

/skills browses skills the agent has either learned or has access to from the broader Hermes skills hub.

Ctrl+C or just typing a new message interrupts whatever the agent is currently working on — this is critical when an agent goes down a wrong path and you want to redirect it without nuking the session.

Configuring Messaging Gateways

The thing that makes Hermes genuinely useful for me is that it lives in my messaging apps, not in a terminal window I have to remember to open.

You start the gateway with one command.

hermes gateway

That spins up the messaging gateway and from a single config you can attach Telegram, Discord, Slack, WhatsApp, Signal, and Email.

For Telegram, you'll need a bot token from BotFather which takes about ninety seconds to create.

For Discord, you create a bot in the Discord developer portal and invite it to your server.

For WhatsApp and Signal, the setup is a little fiddlier because you're bridging to phone-native messaging — follow the gateway prompts and budget ten minutes the first time.

Once it's running, you can voice-memo Hermes from your phone and it transcribes, processes, and replies — that single feature has replaced about half my note-taking workflow.

Watch The Q&A Walkthrough

The Q&A covers most of the install edge cases I get asked about — well worth twenty minutes if you're stuck on a specific install error.

Troubleshooting The Most Common Install Issues

Mac Mini owners regularly hit performance issues running Hermes locally and the cleanest fix is to push Hermes onto a cheap cloud VPS like Hostinger and SSH into it.

If your agent feels passive or non-autonomous, the issue is almost always the SOUL.md file — ask Hermes "where is my soul.md file?" and then edit it directly with instructions like "be more proactive, be more autonomous, take action without asking for confirmation."

If you want a one-off persistent loop where the agent keeps working until a goal is hit, use /goal <text> which fires up a goal with a 20-turn default budget.

If your Windows install is failing, the cleanest workaround is to use Claude Code locally — paste the GitHub install instructions into Claude Code and let it iterate through the install for you.

If you can't get a specific model provider working, run hermes doctor and it diagnoses most config and credential issues in one pass.

Migrating From OpenClaw To Hermes

If you're already running OpenClaw and want to switch — and most serious users have switched at this point — Hermes ships with a one-command migration.

hermes claw migrate

That's the full interactive migration with the full preset.

hermes claw migrate --dry-run previews exactly what would be migrated without writing any files, which is what I recommend running first.

hermes claw migrate --preset user-data migrates without secrets if you want to keep your API keys separate.

hermes claw migrate --overwrite overwrites existing conflicts on the Hermes side — only use this if you're sure you don't want to keep any pre-existing Hermes config.

What gets imported includes your SOUL.md persona, MEMORY.md and USER.md entries, user-created skills (pushed to ~/.hermes/skills/openclaw-imports/), command allowlist, messaging settings, allowlisted API keys for Telegram, OpenRouter, OpenAI, Anthropic, and ElevenLabs, TTS workspace audio files, and AGENTS.md workspace instructions.

It's the cleanest migration tool I've seen in the AI agent space.

What Makes v0.13.0 Special

The current latest release as of May 2026 is v0.13.0, codename Tenacity, which dropped on 7 May.

That's eleven releases in the last six months which tells you everything you need to know about the project velocity.

v0.13.0 ships harder retry semantics, better skill curation, faster gateway reconnects, and improved memory consolidation when sessions get long.

If you're running anything older than v0.12, just run hermes update and grab the latest.

Hermes Setup Methods Compared

Install Method	Best For	Setup Time	Persistence
Linux/macOS/WSL2 (bash)	Power users on Unix-like systems	5 min	Local disk
Windows PowerShell	Native Windows developers	7-10 min	Local disk
Termux (Android)	Mobile / travel agents	10 min	Local phone storage
Docker	Reproducible, sandboxed runs	5 min	Container volume
Modal / Daytona serverless	Always-on agents at near-zero idle cost	15 min	Serverless persistent
SSH / VPS	Remote always-on box	10 min	Remote disk

For most readers, the bash one-liner is the right starting point — then graduate to Modal or a VPS once you want the agent always on.

🚀 Want hands-on Hermes coaching? AI Profit Boardroom members get daily Q&A with me and weekly live calls on Hermes setups. → Join here

How Hermes Pairs With Workspace, Swarm, And Goals

Hermes by itself is already powerful, but the real leverage is pairing it with the Workspace, Swarm, and Goal features built on top.

See my Hermes Agent Swarm post for how I run 17+ agents in parallel for content production and research.

See Hermes Agent Goals for the autonomous loops that keep agents working overnight on multi-step tasks.

See Hermes Agent HUD UI for the visual interface I use when I want to manage multiple agents at once without bouncing between terminal panes.

And see Hermes AI Agent Framework 2026 for the higher-level framework view if you're trying to decide whether Hermes is the right base layer for your stack.

FAQ — Hermes Agent Installation Guide 2026

How long does Hermes installation take?

About five minutes for the actual install and another five minutes to pick a model, run hermes setup, and configure your first gateway.

Is Hermes free to install?

Yes — Hermes itself is free and open-source. You pay for the LLM provider you use through it but the agent layer is zero cost.

What's the cheapest way to run Hermes always-on?

A five-dollar Hostinger or DigitalOcean VPS handles it comfortably, or Modal serverless if you want it to spin up only when needed.

Can I run Hermes on a Mac Mini?

Yes but performance can be sluggish — if you hit issues, push it to a cheap VPS and SSH in instead.

Does Hermes work on Windows natively?

Yes, via the PowerShell installer — currently early beta but functional. If you hit issues, run via WSL2 or use Claude Code to debug the install.

What if I'm coming from OpenClaw?

Run hermes claw migrate and it imports your persona, memory, skills, settings, and API keys in one command.

How do I update Hermes?

hermes update pulls the latest release. Run this monthly because the project ships fast.

Should I upgrade to AI Profit Boardroom for the full course?

If you want the structured installation course plus Workspace, Swarm, Goal, and gateway trainings — plus weekly live coaching — yes. The 7-day refund and 30-day ROI guarantee make it risk-free.

Latest Updates

Hermes Agent Swarm — running 17+ parallel agents on top of your fresh install.
Hermes Agent Goals — autonomous overnight loops once you're set up.
Sonnet 4.8 Review — the model I default to inside Hermes for reasoning work.