hermes-agent logo
hermes-agent
private single-user AI assistant

A personal AI assistant, run by one person, for one person.

hermes-agent is a private instance of the open-source Hermes Agent framework, operated by Chris Prewitt on hardware he controls. It is not a public product, it is not available for sign-up, and it has exactly one user: its owner.

What it does

hermes-agent is an automation and research assistant. When connected to a Google account it uses Google APIs to:

Who can use it

Only the owner. This is a single-user tool running on a private machine. There is no multi-tenant service, no public endpoint, and no way for other people to authorize it against their own accounts.

Privacy

How Google user data is accessed, stored and used is set out in the privacy policy. In short: data stays on the owner's own hardware, nothing is sold or used for advertising, no third party sees it beyond the language-model API used to summarise it, and access can be revoked at any time.

Data access is granted through Google OAuth 2.0 and can be revoked at myaccount.google.com/permissions.

Technical notes

The agent runs locally, talks to Google APIs over HTTPS with an OAuth 2.0 desktop client, and stores its token with owner-only file permissions. Scheduled briefings run on a cron schedule; everything it does is read-only by default, and any outbound mail requires explicit owner approval.

Contact

Questions about this application: christopherprewitt@gmail.com