Video: "NEW Hermes AI Email Agent is INSANE!" by Julian Goldie on YouTube.
What the email integration actually does
Hermes Agent now connects to your inbox via standard IMAP and SMTP — the same protocols your email client uses. Once configured, it polls for new messages at a set interval (15 seconds by default) and processes each one it's allowed to see. It reads the thread, applies whatever context it carries in memory, and sends a reply from the same address.
That last part is important. The replies don't go out from a separate notification address or a bot label — they come from the agent's own configured email account, in-thread, with proper subject lines. From the recipient's perspective, it reads like a normal reply. The difference is that the agent can handle that at 3am on a Tuesday without anyone opening a laptop.
What you actually need to set it up
The setup is genuinely accessible, but it's not a two-click process. You need: a dedicated email address for the agent (something like agent@yourdomain.com — not your personal inbox), the IMAP and SMTP host details for your email provider, an app password rather than your main login password, and a list of email addresses the agent is allowed to respond to.
That last item — the allowed senders list — is the most important configuration decision you'll make. Without it, your agent will attempt to process every incoming email, which includes marketing, automated notifications, and anything else that lands in the inbox. Worth setting up carefully before you let it run unsupervised.
The bit most walkthroughs skip
Using a dedicated account rather than a personal or business inbox isn't just good practice — it's the only safe way to run this. The agent has full access to everything in that inbox via IMAP, so you're handing over read and write access to all incoming mail. A dedicated address keeps the blast radius contained if something goes wrong.
In practice, this means the setup works best when you route specific enquiry types into the agent's address. New contact form submissions, support requests from a particular tier of customer, internal task notifications — those are the right candidates. Not your general catch-all inbox, and certainly not anything that includes sensitive financial correspondence.
What changes in a real business workflow
The obvious use case is out-of-hours response. If someone submits a query at 9pm, the agent can acknowledge it, ask clarifying questions, and pull in relevant context before a human picks it up the next morning. That's not replacing the human response — it's doing the triage so the human response is more useful when it comes.
The less obvious use case is internal routing. An agent that monitors a shared inbox can categorise and forward based on content, track follow-ups it was asked to handle, and flag messages that need human attention — without any of that work sitting in a queue waiting for someone to open their laptop. For small teams that run partly on email, this is genuinely useful. It's not magic, but it removes a category of low-value administrative work that tends to pile up.
Where this connects to NordSys
Hermes Agent is open source and free to install, but the configuration decisions — particularly around allowed senders, memory setup, and model choice — make a significant difference to how useful the email integration actually is in practice. We set up Hermes Agent for UK businesses and include email configuration as part of the standard install, so the agent is doing useful work from day one rather than sitting idle until someone figures out the IMAP settings.
See our AI Agents service →