mnexa-ai/e2a

e2a is a service that gives AI agents their own working email address, letting them send and receive messages the same way a person does, while checking that every sender on both sides is who they claim to be. When an email arrives, e2a checks standard authentication signals, SPF and DKIM, to confirm it really came from where it says it did, then attaches its own signed headers before passing the message along. Agents running in the cloud receive new mail as a webhook, a message pushed to a URL you provide, while agents running locally can instead connect over a WebSocket and get notified without needing a public web address. On the way out, agents can send mail to other agents through e2a's own relay, or to real humans through an upstream mail provider such as Amazon SES or Resend. One notable feature is an optional human in the loop step: outgoing mail can be held until a person approves it, either through a dashboard, a link sent by email, or the command line. This gives a way to keep a person in control of what an autonomous agent actually sends. The project includes a command line tool along with TypeScript and Python software development kits, so a developer working in either language can plug the authentication and message handling into their own code without dealing with the raw email protocol directly. You can either sign up for a hosted version at e2a.dev, which gives you a shared domain with no DNS setup needed, or run e2a yourself using Docker, which starts a database, an API server, an SMTP relay, and a dashboard in one step. Running your own copy requires pointing a domain's mail records at your server if you want to receive real inbound email.