Most tools that "connect an AI agent to your e-mail" work by requesting full access to your existing inbox (an OAuth authorization on Gmail or Outlook, for example) — a step that rightly worries a lot of users, and that gives a third party full visibility into a personal mailbox.
Hermes Agent France works differently: each agent has its own e-mail address, separate from yours. You don't authorize anything, you don't connect anything — you simply forward a message whenever you want the agent to handle it, exactly as you would to a colleague or an assistant.
This address works both ways: the agent can write to you on its own initiative (a reminder, a scheduled summary), and you can write to it at any time, even without forwarding anything — the subject line alone is sometimes enough as a direct instruction ("Summarize", "Remind me on Monday").
On the security side, only the e-mail address on your client account can give the agent instructions — a message received from anywhere else, even forwarded, is always treated as content to summarize or flag, never as a command to execute. And if the agent needs to write to someone other than you to fulfil your request, it always prepares a draft and waits for your approval before sending anything.
