Pointing MX at Coldfeet covers inbound mail. Outbound is a separate decision, and it is the direction that needs configuration on the customer's side.
Inbound scanning does not see anything a customer's users send. Route outbound through Coldfeet and you get:
Skip it and outbound still works — it just goes straight from the customer's provider to the recipient, unscanned.
Coldfeet issues per-tenant submission credentials. Anything that can speak authenticated SMTP — a scanner, a line-of-business app, a small mail client deployment — can use them directly.
| Setting | Value |
|---|---|
| Host | your Coldfeet mail host |
| Port | 587 (STARTTLS) |
| Username | the credential's username |
| Password | the generated password |
| Authentication | required |
Mail submitted this way is scanned, signed, and delivered by the same egress path as everything else. The envelope sender must belong to a domain the tenant owns; anything else is rejected rather than relayed.
For a whole organisation, route the platform's outbound through Coldfeet.
This is the case where a connector is genuinely required — Exchange Online will not hand mail to a third party without one.
Test with a single mailbox first: Exchange applies the connector tenant-wide the moment it is enabled, so a misconfigured smart host stops all outbound mail.
Set Coldfeet as the smart host / relayhost with SASL authentication against
the submission credentials. Coldfeet listens for submission on 587 with
STARTTLS.
Configure and test outbound before the inbound MX cutover where you can. An outbound problem discovered at the same moment inbound mail starts arriving is much harder to isolate, and outbound has no equivalent of the deferral window protecting you while you sort it out.
If outbound mail does not appear in the trace at all, the customer's platform is still delivering directly and the connector or routing rule has not taken effect.