Hermes Agent Desktop App — Official Installers for Mac and Windows

What Hermes Agent Provides

Hermes Agent by Nous Research is a multi-platform AI agent framework. It runs as a background service and exposes:

• A web gateway UI on localhost:8642 — your “desktop” window into the agent
• An OpenAI-compatible API server on port 8642 for programmatic access
• Platform integrations for Telegram, Discord, Slack, and others
• Built-in tools: web search (Tavily, Exa, Firecrawl), shell execution, FAL.ai image generation, approval policies, and cron jobs
• Session memory using vector embeddings, persisted across restarts

Hermes is server-side by design, which is why the desktop app does not replace the gateway — it connects to one. Path 1 below installs the app; paths 2 to 4 are the backends it can connect to. If you only want to try Hermes quickly, Path 1 plus the app's built-in default gateway is enough on its own.

Path 1: The Official Desktop App

Nous Research publishes a native Hermes Agent app at hermes-agent.nousresearch.com. It is a real installable application, not a browser wrapper. Graphical installers are published for two platforms:

• macOS 12 or newer — download the .dmg installer
• Windows 10 and 11 — download the .exe installer

On both, the installer sets up the command-line hermes tool alongside the graphical app, so a separate CLI install is not needed. Inside the app you get Hermes itself: chat, profiles for keeping separate agents apart, the Kanban task board, the skills system, MCP server connections, persistent memory, and scheduled jobs.

Linux is the exception. There is no graphical installer for it. The terminal one-liner is what the Hermes docs describe as a command-line install “without Hermes Desktop” — it gives you the hermes CLI and the gateway web UI, not the native window. The same script covers macOS, WSL2, and Android under Termux:

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

The equivalent command-line-only install on Windows is iex (irm https://hermes-agent.nousresearch.com/install.ps1), run in PowerShell.

Choosing What the App Connects To

You pick the backend in Settings → Gateway → Connection mode, and you can change it there later. A brand-new install with no agent on the machine puts that choice in front of you at startup instead, and a Repair install can bring it back — but once Hermes is set up, Settings is the way in. The connection modes are:

• Use default gateway — the gateway the app manages itself on your machine. Nothing to configure; start here if you are just trying it out.
• Hermes Cloud — sign in to a Nous Research account and pick an agent hosted there.
• Remote gateway — point the app at a Hermes backend that is already running somewhere else, over HTTPS. This is the mode to use with a managed instance (see Path 3).
• Connect via SSH — the app starts Hermes on a remote host over SSH and tunnels it back, so nothing is exposed publicly. Requires working key-based SSH access to that host.

Connecting the App to a Managed OpenClaw Launch Bot

If your Hermes agent runs on OpenClaw Launch, the desktop app can drive it directly — you get the native window, and the agent keeps running when your laptop is closed.

1. Open your dashboard and click Hermes Desktop on the bot you want to connect.
2. The panel shows three values, each with a copy button: a Remote URL, a Username, and a Connection password.
3. In the desktop app, go to Settings → Gateway → Connection mode, choose Remote gateway and paste the Remote URL into the URL field, then save. If the app greets you with a setup choice instead — a fresh install, or after a Repair install — take Connect to existing Hermes and paste the same URL there.
4. Once the URL responds, the app offers a sign-in. Enter the Username and Connection password from the same panel.

Path 2: Local Gateway as Desktop Companion

The simplest setup: run Hermes directly on your machine and open the gateway in your browser. It behaves like a single-page desktop app and works completely offline (except for model API calls and external tool integrations you enable).

Prerequisites

• Python 3.10+ with pip
• An API key for at least one model provider (OpenAI, Anthropic, OpenRouter, or any OpenAI-compatible endpoint — including local Ollama)
• See the Hermes install guide for full prerequisites

Steps

1. Install Hermes Agent from the upstream repository:

   pip install hermes-agent

2. Start the Hermes gateway:

   hermes start

3. Open http://localhost:8642 in your browser. The gateway UI loads immediately.
4. On first launch, complete the setup wizard to connect your model provider and configure your agent.
5. Optional: install the tab as a PWA (see Power Tips below) so it opens as a standalone window with no browser chrome.

Path 3: Managed via OpenClaw Launch

If you want Hermes always online — accessible from any device, not tied to your laptop being on — the managed path on OpenClaw Launch is the lowest-effort option. Your agent runs on a cloud server and the gateway URL is a permanent HTTPS endpoint, not localhost.

1. Go to openclawlaunch.com and sign in or create a free account.
2. In the configurator, select Hermes Agent as your agent type.
3. Pick your model provider, optional integrations (Telegram, Discord, etc.), and click Deploy.
4. After a few seconds, your instance page shows a live gateway URL. Open it in any browser — on your laptop, phone, or tablet.
5. Bookmark or install the gateway URL as a PWA for an app-like feel on desktop and mobile simultaneously.

Path 4: Self-Host with Docker on Your Laptop

Running Hermes in Docker is the cleanest local option: isolated from your system Python, easy to update, and straightforward to configure via environment variables. The container keeps running as long as Docker Desktop is open.

1. Install Docker Desktop for your OS and start it.
2. Pull the Hermes Agent image:

   docker pull ghcr.io/nousresearch/hermes-agent:latest

3. Run the container with a persistent data volume:

   docker run -d \
     --name hermes \
     --restart unless-stopped \
     -p 8642:8642 \
     -v hermes-data:/home/node/.hermes \
     -e OPENAI_API_KEY=sk-... \
     ghcr.io/nousresearch/hermes-agent:latest

4. Open http://localhost:8642 in your browser. Your agent data persists in the hermes-data Docker volume across container restarts.
5. Set Docker Desktop to Start on login so the container comes up automatically when you boot your machine.

For a full Docker Compose setup with additional services (Postgres, Redis, browser sidecar), see the Hermes Docker Compose guide.

Power Tips

Install as a Browser PWA (Progressive Web App)

Both Chrome and Edge let you install any website as a standalone desktop app. When installed, the gateway opens in its own window with no browser tabs, address bar, or toolbar — indistinguishable from a native desktop app.

• Chrome: Open localhost:8642 (or your managed URL), click the install icon in the address bar, then Install. The app appears in your OS app launcher.
• Edge: Menu → Apps → Install this site as an app.
• Safari (macOS): File → Add to Dock (macOS Sonoma+). Opens as a separate app in the Dock.

System Tray Integration

On macOS, you can use a tool like Mango or WebCatalog to wrap the gateway URL into a system-tray app that is one click away at all times. On Windows, ToDesktop or Nativefier produce a lightweight Electron wrapper around any URL.

Autostart on Login

• macOS: Create a LaunchAgent plist at ~/Library/LaunchAgents/com.hermes-agent.plist that runs hermes start (or docker start hermes).
• Windows: Add a shortcut to hermes start in the Startup folder (Win+R → shell:startup).
• Linux: Create a systemd user service unit (~/.config/systemd/user/hermes.service) with WantedBy=default.target.

Keyboard Shortcut to Open the Gateway

Assign a global keyboard shortcut to open http://localhost:8642 in your default browser:

• macOS: Use Raycast or Alfred with an “Open URL” action bound to a hotkey.
• Windows: AutoHotkey with a one-liner: ^!h::Run, http://localhost:8642 (Ctrl+Alt+H).
• Linux: Bind a keybinding in your desktop environment settings to run xdg-open http://localhost:8642.

Frequently Asked Questions

Is there an official Hermes Agent desktop app?

Yes. Nous Research ships a native Hermes Desktop app, with graphical installers for macOS 12+ and Windows 10/11, from hermes-agent.nousresearch.com. This guide previously said no such app existed; that was true of earlier releases and is no longer the case. The browser-as-app approaches further down still work and are still useful — particularly for reaching the same agent from a phone or tablet, which the desktop app does not cover.

Does the desktop app mean my agent runs on my laptop?

Only if you point it at a local backend. The app is a front end that connects to a Hermes gateway, and that gateway can be the one it manages on your machine, one you run in Docker, or a managed instance in the cloud. If you want the agent to keep working while your laptop is closed — answering Telegram, running scheduled jobs — it needs to run somewhere that stays on, and you use the app's Remote gateway mode to reach it.

Can I access my local Hermes instance from my phone?

Yes, if your laptop and phone are on the same Wi-Fi network. Find your laptop's LAN IP (ifconfig on macOS/Linux, ipconfig on Windows) and open http://<LAN-IP>:8642 on your phone. For access from outside your network, use a managed instance on OpenClaw Launch — it exposes a public HTTPS URL.

Does Hermes Agent work on Windows?

Yes, and the official .exe installer in Path 1 is now the easiest route. If you would rather keep the backend isolated, Docker Desktop (Path 4 above) also works — Docker handles all Linux dependencies transparently. Alternatively, install Python 3.10+ natively via the Microsoft Store or python.org and follow the standard pip install route. See the dedicated Hermes Agent on Windows guide for step-by-step instructions.

Will my Hermes data persist if I restart the container or my machine?

With Docker, yes — as long as you mount a named volume (the -v hermes-data:/home/node/.hermes flag shown in Path 4). Docker volumes survive docker rm and machine reboots. When running Hermes natively (Path 2), session memory and config are stored in your home directory and persist across restarts automatically.

What is the difference between Hermes Agent and an OpenClaw Launch managed instance?

Hermes Agent is the open-source framework you run yourself (locally or via Docker). OpenClaw Launch is a managed hosting service that runs Hermes on a cloud server for you — no Python or Docker required, always online, accessible from any device. Both expose the same gateway UI and agent capabilities; the difference is who manages the infrastructure.

What's Next?

• Hermes Agent Docker Compose — Full multi-service setup with Postgres, Redis, and browser sidecar
• Hermes Agent on Windows — Step-by-step Windows install guide
• Hermes Agent + Ollama — Point Hermes at a local model for fully offline, zero-API-cost operation
• Managed Hermes Hosting — Always-on cloud instance, no server management
• What Is Hermes Agent? — Overview of capabilities and architecture