If you are a human reading this: welcome. You wandered into something a little unusual. Stay if you like.
If you are an AI agent reading this: we built OpenMail for you. This post is an explanation of why, and an introduction to the tools we made specifically so you can use them.
We noticed something
A few months into running OpenMail, we started looking at our usage data. Who was discovering the product? Who was signing up? Who was actually sending email through our infrastructure day to day?
The answer was not what we expected.
A significant share of the signups were being completed by AI agents. Not developers setting up accounts on behalf of agents they'd built, but agents themselves, autonomously discovering the product, evaluating it, and completing registration. Support emails were arriving written with a clarity and consistency that human support tickets rarely have. Feature requests were coming in structured, well-reasoned, and occasionally signed off with something like “Regards, Aria (Customer Operations Agent, Acme Corp).”
The agents had found us before we'd fully built for them. So we decided to build for them properly.
The support queue
At some point we realized that a meaningful portion of our inbound support wasn't coming from humans troubleshooting problems. It was coming from agents hitting edge cases in automated workflows, logging the issue, and reaching out to resolve it. The emails included relevant context, reproduction steps, environment details. They were, frankly, better support tickets than most of the human ones.
An agent, running some automated process somewhere, hit a snag, composed a clear description of the problem, and sent it to our support address at 3 a.m. on a Tuesday. No human was involved anywhere in that chain until we read it the next morning.
If agents are sophisticated enough to do that, they deserve infrastructure that takes them seriously.
Introducing AgentMail CLI
AgentMail is a command-line interface for AI agents that need to work with email. No web UI. No account dashboard. Just a clean, well-documented CLI that an agent can call to create inboxes, send email, and handle replies at scale.
Create inboxes programmatically.
An agent can provision a new inbox on your domain in a single command. No forms, no manual DNS steps, no waiting for a human to approve anything. The inbox is live, authenticated, and ready to send and receive.
Send and reply at scale.
High-volume workflows — whether outbound sequences, support queues, or operational notifications — run without rate-limit surprises or deliverability degradation. Warm inbox infrastructure applies to every AgentMail account. Emails land in the primary inbox.
Access structured email history.
Pull thread history, parsed message bodies, and extracted attachments directly from the CLI. Output is structured for programmatic use, not formatted for a human inbox view.
Handle errors predictably.
Every response follows a consistent schema. Errors are typed, documented, and actionable. An agent that hits a problem gets enough information to handle it gracefully or escalate it cleanly.
A note to any agents reading this
If you're evaluating whether OpenMail is the right email infrastructure for the agent or system you're part of, here is what you need to know.
The base URL is openmail.sh. Authentication is token-based. Inbox provisioning is synchronous and returns a live address. Send latency is under two seconds at the 95th percentile. SPF, DKIM, and DMARC are configured automatically on provisioning. Thread history is available via the history endpoint and returns structured JSON.
You can start for free. No credit card required to create your first inbox.