Hermes Agent Declares "Fully Open to Forking and Self-Hosting"

On August 17, Nous Research added a "bot mode" to its desktop app, Hermes Desktop, allowing an agent profile to be split into multiple role-specific bots. Three days later, on August 20, it introduced an option to attach a dedicated remote computer, letting the agent run 24/7 for a few dollars a month. This latest post, published the same day, reads less like a feature-specific update and more like a statement of the overall operating philosophy behind Hermes Agent.

"Fork It, Run It From Your Fridge"

In the post, Nous Research stated that there are no restrictions on modifying, extending, sharing, or forking Hermes Agent. It can be used through the desktop app or a terminal interface (TUI), and can also be messaged via text. The company emphasized the breadth of possible usage environments with the phrase: "talk to it, or chat with it from your smart fridge."

The way models are run isn't fixed to a single method either. Users can plug in their own API key, go through the Nous Portal, or run local models on their own infrastructure. The company explained that users can run a single model or as many as fifteen at once, and can even mix responses from multiple models using a Mixture of Agents (MoA) approach. It also cited a wide range of hardware examples, from workstations like Mac Studio to $35 single-board computers (SBCs), hand-me-down ThinkPads, Docker containers, and Hermes Cloud VPS, which it said costs about as much as a cup of coffee per month. The software itself is open-sourced at the github.com/NousResearch/hermes-agent repository.

An Independent Guide Site Also Emerged

Around the same time, a site called usehermesagent.com appeared, positioning itself around installation and operation guides for Hermes Agent. The site states upfront on its landing page that it is "not owned or endorsed by Nous Research," making clear it is not an official channel. While the official site explains the product itself, this site describes its own focus as gathering practical information—installation scripts, memory configuration, skills, tool gateways, and comparisons of cloud hosting options.

According to the site, Hermes Agent operates like an always-on operator residing in channels such as Slack, Discord, Telegram, WhatsApp, Signal, email, and CLI. It also describes a memory feature that saves repetitive tasks as skills to be recalled in later sessions, a sub-agent delegation structure that handles research, code review, and browser QA simultaneously in separate contexts, and a tool gateway that combines web search, browser automation, image generation, and voice conversion. The site also states that it is preparing content comparing Hermes Agent with competing agents such as OpenClaw and Claude Code.

How to Try It

Setting up Hermes Agent yourself involves three steps.

1. Install — Run curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash in a terminal to download the official open-source runtime.
2. Configure — Run hermes setup to choose which model to use and connect channels (Slack, Discord, email, etc.), tools, and memory.
3. Host (optional) — If you'd rather not manage a server yourself, run cloud.hermeshq.ai deploy to deploy to the cloud and keep it running 24/7.

Local installation is available on Mac, Linux, and Docker environments, and the open-source code can be downloaded by anyone from the github.com/NousResearch/hermes-agent repository. For those who just want always-on operation without managing a server, there's also the option of joining the cloud hosting waitlist.

The use cases Nous Research cited in the post are concrete: parsing WhatsApp backups to summarize content, periodically scanning group chats to send alerts when overdue deadlines are mentioned, and scanning weekly news to check the outcome of a bet made with a sibling.

Local vs. Cloud: What's the Difference

Item	Local Execution	Cloud Execution
Example environments	Mac Studio, $35 SBC, ThinkPad, Docker	Hermes Cloud VPS, cloud.hermeshq.ai
Management burden	User manages server, SSL, and background tasks directly	Managed infrastructure handles server setup and gateway maintenance
Cost sense	Hardware purchase and electricity-level costs	About the cost of a cup of coffee per month (per company)
Model operation	Freely combine own keys and local models	Simplified model access via Nous Portal, etc.

Editor's View

What matters here is that this post isn't announcing a single new feature. Nous Research is essentially packaging the message itself—"our agent locks you into nothing"—as a product. Where ChatGPT and Claude's desktop apps start from a single fixed window and gradually bolt on features, Hermes Agent instead hands over full control of the execution environment and model choice to the user from the outset. Publishing an open-source repository and supporting both local and cloud deployment puts it on a fundamentally different footing from commercial agents that have settled into closed subscription models.

The emergence of an independent guide site is also worth noting. The fact that installation and comparison content not produced by the company itself is accumulating organically signals that the open-source project's user base has grown fairly wide. That said, since such sites are operated by parties separate from the company, it's worth verifying trust with the site operator before entering email or payment information into a cloud hosting waitlist.

For teams in Korea considering self-hosting an in-house work agent, the ability to get started with a single installation script script is a clear reduction in the barrier to entry. However, features like memory, tool gateways, and sub-agents only reveal their true value once actual operational data accumulates, so it would be safer to start with low-risk tasks—like email sorting or group chat monitoring—where failure carries little cost. In the coming weeks, it seems likely that Hermes Cloud will roll out an official pricing structure, and that a few more third-party hosting services beyond usehermesagent.com will emerge on the back of it.