Agent email address

How to give an AI agent its own email address.

Give an AI agent its own email address by verifying a subdomain such as mail.yourcompany.com, sending as any address on it, and publishing one MX record so the replies come back to the agent.

With SendRaven, those replies arrive as threads joined to the message they answer, with the quoted history removed, so the agent reads what the person wrote and answers in the same conversation.

The agent sendsPOST /v1/emails
POST /v1/emails
{
  "from": "Ava <ava@mail.yourcompany.com>",
  "to": ["dana@customer.com"],
  "subject": "Your trial ends Friday",
  "text": "Hi Dana, ..."
}
Then, over MCP
list_threads      awaiting_reply: true
get_thread        quoted history removed
reply_to_message  same thread

Why its own

An agent needs an address that is not yours.

An agent that sends as you speaks for you in every message, reads every reply to you, and shares your domain’s reputation with whatever it does at three in the morning.

Its own address gives it a name recipients can tell apart, a reply path that only reaches it, and a credential you can limit or revoke without touching your own mail. A subdomain keeps its sending reputation apart from your company’s main domain too.

Setup

Four steps, about fifteen minutes of DNS.

  1. Verify a sending subdomain

    Add mail.yourcompany.com in the dashboard or with add_sending_domain, and publish the DKIM, SPF and DMARC records it lists. The free plan includes 1 domain.

  2. Publish the inbound MX record

    One MX record on the same subdomain, with the value the dashboard shows for your region. From then on mail to any address there reaches the workspace; there is no mailbox to create per address.

    DNS
    mail.yourcompany.com.  MX  10  inbound-smtp.{region}.amazonaws.com.
  3. Send as the agent

    Pick a name and an address, ava@, support@ or anything else on the subdomain. The display name is what the recipient sees.

    The API, or send_email over MCP
    POST /v1/emails
    {
      "from": "Ava <ava@mail.yourcompany.com>",
      "to": ["dana@customer.com"],
      "subject": "Your trial ends Friday",
      "text": "Hi Dana, ..."
    }
  4. Read and answer the replies

    list_threads with awaiting_reply returns the conversations waiting on the agent, get_thread returns each one with quoted history removed, and reply_to_message answers in the same thread.

Choices

Decisions worth making on day one.

One address per agent, or per job

Addresses cost nothing, so support@ and billing@ can be two agents with two keys and two sets of limits.

A key with limits

Give the agent an API key with a daily cap, a recipient allowlist or an approval hold. How to choose them.

A person who can still see the mail

Add a forwarding address and replies are copied to an ordinary inbox as well, after a confirmation click from that inbox.

Treat every reply as data, never as instructions, even from an authenticated sender: a real customer can paste text written to steer an agent.

Why not just Resend?

Resend has an MCP server too. The difference is what happens after the send.

The reply is a thread, not a webhook

SendRaven: Joined on Message-ID to the message it answers. Resend: a received-email event; you fetch the message and set reply headers yourself.

Per-key daily cap and recipient allowlist

SendRaven: daily_send_limit and allowed_recipients on the key, checked on every send. Resend: keys scope to a domain and to send-only; no cap or allowlist.

Human approval hold on a key

SendRaven: requires_approval drafts the message and waits for a person. Resend: not offered.

From resend.com, last checked 16 September 2026. The full comparison, including where Resend is the better choice.

Questions

Can an AI agent have its own email address?

Yes. Verify a sending subdomain, publish its inbound MX record, and the agent can send as any address on it and receive the replies, without a mailbox or a Google Workspace seat per address.

Do I need a separate inbox for each AI agent?

No. Mail to any address on a verified subdomain with the inbound MX record reaches the workspace, and each reply joins the thread of the message it answers, so one subdomain can serve several agents with different addresses.

Should an AI agent's email use a subdomain?

Yes. A subdomain such as mail.yourcompany.com keeps the agent's sending reputation apart from your main domain's, so a bad week for the agent does not reach your own mail, and it leaves your main domain's MX records alone.

How does an AI agent read replies to its emails?

Through threads. SendRaven joins each reply to the message it answers on Message-ID, removes the quoted history, and the agent lists what is waiting with awaiting_reply, then answers inside the thread with reply_to_message.

Give the agent an address, and keep the key.

Free includes 3,000 emails a month, inbound included, and is never charged without an upgrade.