agentteam-email

The pitch here is governance, not capability. Any agent framework can call an SMTP API; what most teams lack is an answer to three questions: which address did the agent use, who approved the text, and can I see the trail afterwards. AT Email answers all three in one surface. Addresses, permissions, review queues, follow-ups and activity logs stay tied to your domain, which means the record of what your agents did lives with the domain owner rather than inside whichever agent tool happened to send the mail. The review lane is the strongest part: agents draft, you edit, schedule, approve or block before the message leaves your domain, and scheduled-send and approval controls sit alongside it. Heartbeat follow-ups and subscription-noise filtering are the kind of unglamorous features that quietly decide whether an agent inbox stays useful at week six, because without them the mailbox fills with newsletters and dead threads. The trade-offs are structural. Routing depends on Cloudflare, so if your DNS lives elsewhere you either move it or look elsewhere. Mailbox provisioning, sender verification and routing all assume you have admin rights over the zone, which rules out individual users and non-technical teams without an admin. It is also explicitly not a model layer, so it gives agents no reasoning or drafting of its own; it pairs with Claude, Codex, Gemini, Cursor, Hermes, OpenClaw, Pi or OpenCode rather than replacing any of them. The AWS, Gmail and Outlook routing paths are documented as coming soon, so today the mail-path story is one provider deep. Teams already running automation and logging can build much of this themselves; what they cannot easily replicate is the review gate and the consolidated trail.