3 min read
Nous Research shipped Hermes Desktop yesterday as a public preview, and now the self-improving agent that the AI community has been obsessing over has an official front door. No terminal or config files. Just download, install, and run.
Previously, if you wanted to run Hermes with any kind of visual interface, you had to find it yourself.
The community had built some genuinely impressive options—we covered four of the best ones last month—but all of them were unofficial, third-party GUIs, SSH companions, or browser-based wrappers.
Hermes Agent is an open-source autonomous AI agent built by Nous Research that does something most AI tools don't: It gets better the more you use it. When it figures out how to do something new, it saves that approach as a reusable skill document so it can apply it again next time, without you having to explain it again. That's the core pitch, and it's why Hermes became so popular.
The catch was always the setup. To run Hermes, you had to open a terminal, run a curl install command, and then type your way through every interaction. That's fine if you're a developer and the terminal is your natural habitat, but it's a real barrier if you're not.
Hermes running via Terminal
OpenClaw, Hermes's main open-source rival, had already solved this. It ships with a visual interface out of the box, which is part of why it built its user base faster among non-technical people early on. Hermes had the more interesting architecture—specifically that self-improving skills loop—but it was also the harder one to actually get running.
The desktop app closes that gap.
Hermes Desktop is a native app for macOS 12+, Windows 10/11, and Linux, built on Electron and React with a Python backend. It runs the same agent core as the CLI version—same memory, same skills, same configuration—so nothing you've built before gets left behind.
The macOS and Windows builds are direct downloads. Linux users still install via terminal, which is a reasonable compromise for that platform.
The feature list covers what you'd expect from a serious agent interface. Persistent memory that stores your projects and how it solved them, natural-language scheduling for recurring tasks and reports, web browsing and image generation, and support for more than 300 models through the Nous Portal.
You can delegate to sub-agents—essentially smaller instances of Hermes that run their own isolated tasks in parallel—and reach the same agent across Telegram, Discord, Slack, WhatsApp, Signal, and email. Five execution backends handle sandboxed environments: local, Docker, SSH, Singularity, and Modal.
The portal has a free tier. Paid plans—Plus, Super, Ultra—come with monthly credits and model access. The app itself is MIT-licensed, meaning anyone can audit the code, self-host it, or modify it without asking permission.
The new app also makes Hermes substantially more approachable than OpenClaw for anyone coming in fresh. OpenClaw is a deep ecosystem with a skill marketplace, 50+ messaging integrations, and a large community. It's powerful and it has its own learning curve. Hermes's pitch has always been simpler to state: use it, and it learns.
Nous Research notes the app is in public preview, which means rough edges should be expected. The GitHub repository is live and the team is actively collecting feedback.
Decrypt-a-cookie
This website or its third-party tools use cookies. Cookie policy By clicking the accept button, you agree to the use of cookies.