Everything your agent can do,
inside what you approve.

A free inbox that takes real mail for an hour, then closes, so a signup cannot follow you.

Every address on the domain receives, including ones that have never been created. One switch closes it.

Reply from the exact address the mail arrived on.

Autosaved three seconds after you stop typing, and reopened from the Drafts folder with the recipients, the subject, the body and the From address intact. The attachments stay behind.

Name a label, pick a colour, and filter the list down to it.

Keep a private note attached to a thread.

Every attachment this mailbox holds, in one place you can search, download and clear out.

Pick them with the paperclip or paste them in. A pasted image lands in the body rather than the file list, and a file dragged onto the composer is not attached at all. On an OpenEmail mailbox incoming files sit in object storage, one object each, and opening a thread pulls every one of them.

Download links for the files you send on a subdomain you choose, such as files.yourdomain.com, instead of the OpenEmail address.

Tell your endpoint when mail arrives, instead of making you poll.

Your own conditions, and what should happen to anything that matches them.

Whether a message you sent has been opened, when, how many times and which of its links were followed. Off until you turn it on, and plain about the readings it cannot get.

Point several domains at one inbox, subdomains included.

A starter p=none record, printed to copy or written by a sync, never tightened for you.

Forgiving words and operators like from: and has:, inside the folder you are reading.

What your mailbox is actually doing, drawn rather than guessed at.

Send one message or a hundred per call, now or later, and retry without sending twice.

A window to change your mind, and the composer comes back exactly as it was.

Write it now, land it when they are awake.

Point Claude, or any MCP client, at your mailbox.

Your mailbox and every API operation, from a terminal or a script.

A documented HTTP API with issuable, scopable, revocable keys.

A TypeScript client first, then the rest.

Apps register themselves and sign in with PKCE, and you choose the workspace, addresses and permissions each one gets.

One address means one address.

A role says what somebody may do, and an address grant says where, unless the role reaches every address.

A key per job, limited to the scopes and senders it needs, with every call logged.