Hermes Agent Verified Customer Reviews (2026)

My rating concerns billing support for the paid Nous Portal service. My original refund request of 18 September 2026 has reference T-4167. The automatic reply to my follow-up of 23 October supplied another reference, T-6537. As of 24 October, I have received automatic acknowledgements but no substantive answer to my refund request.

A further subscription charge was processed on 18 October while the original request remained unresolved. I have requested reimbursement of the charge raised in September and the October renewal charge to the original card, plus an update on my previously submitted account-closure request. I still have no confirmation of a refund or account closure.

I rate this billing and support experience two out of five. The automatic acknowledgements give me references to quote, but more than five weeks without a substantive response is poor support in my experience.

The automatic acknowledgement confirmed receipt of my request and supplied reference T-4167, which I could quote in my follow-up messages.

As of 24 October I have only received automatic acknowledgements under T-4167 and T-6537, despite follow-ups. A further subscription charge was processed on 18 October while the original refund request remained unresolved.

Five stars from me. Hermes Agent feels like one of the more compelling open-source agent projects because it combines memory, skills, scheduling, tools, and cross-platform access in a way that actually fits how developers think about automation.

I would not run it recklessly, and I would definitely lock down permissions. But for developers and AI power users who want a self-hostable agent that can grow with their workflows, Hermes Agent is a very exciting tool to experiment with.

Hermes Agent is really interesting from a developer’s point of view because it feels like a real agent framework instead of another chatbot wrapper. I like that it has persistent memory, reusable skills, tool use, cron jobs, and messaging integrations built around the idea of an agent that keeps improving over time.

The open-source MIT license is a big plus. As a developer, I want to inspect the code, self-host it, modify the runtime, and understand how the agent is making decisions instead of depending completely on a closed platform.

The self-improving skills system is the most exciting part. If Hermes can actually turn repeated workflows into reusable skills, that makes it much more useful for long-term automation, personal agents, dev workflows, research tasks, and ops work.

I also like that it can run across terminal, desktop, cloud, and messaging channels. That makes it feel closer to a persistent digital teammate than a tab I have to keep open.

The power also creates risk. Recent security reporting says Hermes Agent has already been abused in real-world intrusion activity when operators ran it with minimal supervision, which is a reminder that autonomous agents need tight permissions, logging, and human review.

It is also probably not the best starting point for casual users. To get the most out of Hermes, you need to be comfortable with setup, self-hosting decisions, access control, model choices, and deciding exactly what the agent should be allowed to do.

I would be careful with anything involving credentials, production systems, customer data, or destructive actions. Hermes looks powerful, but it should be treated like a junior employee with limited permissions, not like a fully trusted autonomous operator.