Nous Research Ships Hermes Desktop So Open-Source Agents Escape the Terminal

Hermes Agent has lived in the terminal for most of its life. Now Hermes Desktop is out in public preview, giving the open-source agent a proper native window on macOS, Windows, and Linux. It ships bundled with Hermes Agent v0.15.2 and is released under the MIT license across macOS 12+, Windows 10/11, and Linux.

The release was teased at NVIDIA's GTC Taipei keynote, where Jensen Huang demoed Hermes alongside NVIDIA's own agent stack. Build-a-Claw signals how rapidly the developer community and ecosystem are scaling agents. Starting with OpenClaw and Hermes Agent, attendees configured their claw's persona, added agent skills and set its schedule. The Desktop launch turns that demo into something anyone can install.

Same agent, new surface

The key architectural decision is that this is not a fork or a wrapper with its own state. The desktop reuses the same agent core. It shares configuration, API keys, sessions, skills, and memory with the CLI and gateway. The desktop is another surface over one agent, not a fork.

That means a conversation you start in the desktop window resumes verbatim in hermes on the terminal, and vice versa. Under the hood, the packaged app ships only the Electron shell. On first launch it installs the Hermes Agent runtime into HERMES_HOME (~/.hermes, or %LOCALAPPDATA%\hermes on Windows) , the same layout a CLI install uses, so the two are interchangeable. The renderer (React, in src/) talks to a hermes dashboard --tui backend over the standard gateway APIs and reuses the embedded TUI rather than reimplementing chat.

What the window actually gives you

The GUI is built around chat, but it surfaces a lot of agent internals that were previously buried in YAML files and TUI panels:

• A right-hand preview rail that renders web pages, files, and tool outputs side by side while you keep chatting.
• A working-directory file browser, useful for following along as the agent reads, writes, and edits files.
• Voice in and out, the same mode available elsewhere. On macOS the OS will prompt once for microphone access.
• A real UI for managing providers, models, tools, and credentials instead of editing YAML.
• Streaming tool output, so you can watch each tool call land in the chat instead of waiting for the agent to finish.

There is also a remote mode. By default the app starts and manages its own local backend. You can instead point it at a Hermes backend running on another machine , a VPS, a home server, or a Mini behind Tailscale. A username and password protect the dashboard, and the app signs in over the gateway API.

The agent underneath

For anyone who has not followed the project, the thing the desktop is fronting is genuinely unusual. Nous research team describes Hermes as having a closed learning loop. This is what separates it from a simple chat wrapper. After a complex task, the agent writes a reusable skill. Those skills then self-improve during later use. Memory is persistent and agent-curated, with periodic nudges to save knowledge. Cross-session recall uses FTS5 session search with LLM summarization.

It is also model-agnostic. You can use Nous Portal, OpenRouter (200+ models), NovitaAI, NVIDIA NIM (Nemotron), Xiaomi MiMo, z.ai/GLM, Kimi/Moonshot, MiniMax, Hugging Face, OpenAI, or your own endpoint. Switch with hermes model , no code changes, no lock-in. For execution, five backends are available for sandboxed execution: local, Docker, SSH, Singularity, and Modal.

And the agent does not live only inside the window. It runs across platforms including Telegram, Discord, Slack, WhatsApp, Signal, email, and the terminal. Start a task in the desktop preview pane, get the result pinged to you on Signal an hour later.

Install path

macOS and Windows users get a direct installer from the Hermes Desktop site. Linux users go through the terminal:

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

It builds and launches the GUI against your existing install , same config, keys, sessions, and skills. On first launch Hermes walks you through picking a provider and model; nothing else to configure. The source for the desktop app lives in the main Hermes repo.

The catch

This is a public preview, not a 1.0. Early users on social have praised the install experience but reported glitches and setup pain on certain configs. Worth flagging too: self-improving claims are still vendor-stated. The desktop changes the UX, not the model or the security model. Benchmark on your own workloads before trusting the numbers.

Still, the strategic point is hard to miss. Open-source agents have spent the last year fighting to escape the terminal without splintering into incompatible UIs. Hermes Desktop is a bet that the way to do that is one runtime, many surfaces, with the CLI, the desktop, and seven messaging integrations all reading from the same memory store. That is the part worth watching more than the install button.