Nous Research Launches Hermes Desktop: Native macOS, Windows & Linux App for the Self-Improving Hermes Agent

Nous Research just shipped Hermes Desktop — a native, cross-platform front end for the self-improving Hermes Agent. One installer on macOS, Windows, or Linux, and you get the same agent that has been running headless in terminals and on servers, now wrapped in a real UI.

It is not a chat-wrapper. The desktop app loads the same Hermes Agent core, the same config, the same API keys, the same sessions, the same skills, the same memory. Whatever you set up in `hermes` on the CLI shows up here, and anything you do here shows up there. MIT-licensed, open source.

Why this matters

For two years the most capable open-source agents have been gated behind a terminal. That filtered out everyone who is not comfortable editing YAML or piping `--insecure` flags to a dashboard. Hermes Desktop removes that wall without dumbing the agent down — the power user keeps the CLI, the new user gets a real app, and they're literally the same process underneath.

A few things are quietly significant:

Install in under two minutes

Three paths, pick whichever fits:

1. Native installer (recommended for macOS / Windows). Download from hermes-agent.nousresearch.com/desktop. macOS builds are signed and notarized; Windows ships NSIS + MSI installers with checksums. Run it, done.

2. One-line CLI install (Linux / macOS / WSL2). Adds the desktop app on top of a regular Hermes install:

```bash

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

| bash -s -- --include-desktop

```

3. Already running Hermes? Just type:

```bash

hermes desktop

```

It uses your existing config, keys, sessions, and skills. First launch installs the runtime into `~/.hermes` (or `%LOCALAPPDATA%\hermes` on Windows) — the exact same layout a CLI install uses, which is why the two interchange cleanly.

What's in the app

Chat-first window with a left sidebar for navigation. The pieces:

The app checks for updates in the background and offers one-click upgrades. You're not pinning a brittle version.

Connect it to a remote backend

The killer move: point the desktop app at a Hermes backend running somewhere else — a VPS, a home server, a Mac Mini behind Tailscale.

On the remote machine, pin a stable session token and run the dashboard bound to a reachable address:

```bash

TOKEN=$(openssl rand -base64 32)

echo "HERMES_DASHBOARD_SESSION_TOKEN=$TOKEN" >> ~/.hermes/.env

chmod 600 ~/.hermes/.env

echo "$TOKEN" # paste this into the desktop app

hermes dashboard --no-open --insecure --host 0.0.0.0 --port 9119

```

In the app, go to Settings → Gateway → Remote gateway, paste the URL (`http://<host>:9119`) and the token, hit Test remote, then Save and reconnect. The token is stored encrypted in local config.

A warning that matters: `--insecure` exposes a port that can read your `.env` (API keys, secrets) and execute agent commands. Never put it on the open internet. Bind it to a Tailscale IP and only your tailnet can reach it.

The honest benefits

After a week of running it the practical wins shake out as:

Where this fits in the stack

Hermes is increasingly the obvious answer for teams who want an autonomous agent they actually own — running their integrations, their data, their context, on infrastructure they control. The desktop app does not change that thesis; it just removes the last friction in adopting it. If you've been waiting for "the open-source ChatGPT desktop app that is actually an agent and actually yours," it's here.

Download is at hermes-agent.nousresearch.com/desktop. Source on GitHub. Two minutes to your first message.