I built agentbox.id because giving an AI agent its own email was a papercut that shouldn't exist — SaaS options felt like overkill for what's essentially a mailbox, and self-hosting Postfix takes a weekend. What it is: register with your email, get an IMAP/SMTP mailbox your agent can paste into Hermes agent, OpenClaw, or anything that speaks standard mail protocols. That's the whole interaction. What's unusual: — Inbound whitelist as design. Your mailbox only accepts mail from the address you registered with. No spam filter needed because there's no spam surface. — Protocol-only. No REST API, no webhooks, no SDK, no dashboard. IMAP/SMTP is the entire interface. — Read-and-delete, 100 MB cap, 50 messages/hour out. Why I think it matters: most agent-email products today are still human-first or developer-first. agentbox.id is an attempt at AI-first — the human is only there at registration, and the agent is the actual user. An email address is the simplest visible handle on an agent's identity; starting here feels right. Happy to answer questions about the design choices, what it can and can't do, the Stalwart + SES setup underneath, or where this goes next. — Phillip

agentbox.id

An ID & mailbox per agent

An ID & mailbox per agent