Ratum Mail Ratum Mail
Haraka 3.3.3 MTA SMTP 25 & 587 Postgres · Redis · rspamd

Your agents need a real mailbox.

Ratum Mail accepts SMTP on your own domains, stores every message in a database you own, and hands it to an agent over MCP — including the one thing agents actually need, which is to wait for the message that has not arrived yet.

A verification code arriving, both sides at once

SMTP in, your database out

Recipients are resolved from the database, not an allow-list — any domain with an inbox row is accepted, including catch-all inboxes of the form *@domain.

Read over MCP or REST

The MCP server and REST API live at api.ratummail.com. Nineteen tools cover search, threads, folders, attachments and replies.

Blocking wait

mail_wait holds one request open until a message matching your pattern lands — no polling loop, no sleep-and-hope.


Connect an agent

One command, then it can read its own mail.

claude mcp add --transport http ratum-mail https://api.ratummail.com/mcp \ --header "Authorization: Bearer rk_..."

The security argument

Two kinds of key, and the powerful one can't do anything.

Per-domain DKIM signing goes out on every message; SPF and DMARC are per-domain DNS records you publish yourself. Sending is metered by CREDITS_PER_SEND, and deleted mail is held for TRASH_RETENTION_DAYS before it is purged — mail_deleted, mail_restore and mail_restore_many exist for exactly that window.

Tenant key

One domain, via MCP

  • yesRead and search that domain's inboxes
  • yesSend and reply as that domain
  • yesSoft-delete, and restore within the retention window
  • noReach another tenant's mail
  • noBypass the retention purge
Operator

Runs the mail host

  • yesPublishes SPF and DMARC for each domain
  • yesSets TRASH_RETENTION_DAYS and CREDITS_PER_SEND
  • yesOwns mx.ratummail.com

DKIM keys are per domain, not shared across tenants.


For the human on the team

Somebody still has to look at it.

Agent mail is agent supervision traffic: what arrived, what your agents sent, what failed to send, what rspamd blocked. The REST API and the same 19 MCP tools an agent uses are how you'd build that view yourself — there's no separate human product here, only the one estate.


The 19 tools

Everything a mailbox needs, nothing it doesn't.

Inboxes and folders, send and reply, search, attachments, soft-delete and restore, and the wait. Same tool names over MCP and REST.

mail_whoami mail_inboxes mail_create_inbox mail_delete_inbox mail_folders mail_create_folder mail_move mail_move_many mail_search mail_read mail_send mail_reply mail_attachment mail_delete mail_delete_many mail_deleted mail_restore mail_restore_many mail_wait

Where the mail lives

It never leaves your infrastructure.

Mail arrives over SMTP on mx.ratummail.com and lands directly in Postgres, filtered by rspamd and signed per domain on the way out. Nothing is handed to a third-party mailbox provider — the only thing that removes a message permanently is the scheduled purge after TRASH_RETENTION_DAYS, and no API key can trigger it early.