Exchange mail flow guides › Outbound: send connectors and smart hosts

Multiple smart hosts in Exchange 2016/2019: per recipient domain (built in) and per sender (with a relay)

Exchange supports several smart hosts natively only by recipient domain: one send connector per destination address space, each with its own smart host and cost; routing by sender needs an additional relay hop. This guide shows both setups step by step, the commands to verify them, and a table to decide which one your situation calls for.

Updated on 2026-09-17

First decide which “multiple” you mean

“Several smart hosts” describes two different requirements, and only one of them is a checkbox in Exchange.

• Different destinations, different smart hosts. Mail to a partner, a parent company or a particular country must leave through a dedicated gateway, everything else through the normal route. The decision depends on the recipient. Exchange does this with its own send connectors (method A).
• Different senders, different smart hosts. Mail from company-a.example must leave through provider A with login A, mail from company-b.example through provider B with login B, because each provider relays only for its own customers’ addresses. The decision depends on the sender. Exchange has no setting for that; it takes one extra relay hop (method B). Why that is so is the subject of the companion guide Exchange send connectors route by recipient domain, not sender.

How Exchange picks one of several send connectors

Microsoft documents the selection in Connector selection in external message routing. For a message to an external recipient, Exchange first drops every connector whose message size limit is smaller than the message, then keeps only connectors that are enabled, visible from the local Active Directory site (unscoped, or scoped to this site) and configured with an address space that matches the recipient’s address. From that list the connector with the most specific address space wins. Microsoft’s example: for julia@marketing.contoso.com the order of preference is marketing.contoso.com, then *.contoso.com, then *. Only when two connectors match with the same specificity does the cost decide, lower being preferred, followed by the hop count and finally the connector name.

The address space syntax is short (Send connectors in Exchange Server): * means all domains, contoso.com means that domain without its subdomains, *.contoso.com means the domain and all subdomains. In the Exchange Management Shell the full form is Type:AddressSpace;Cost, for example SMTP:*;1; the cost ranges from 1 to 100, and a lower cost is the better route (New-SendConnector).

Method A: one smart host per recipient domain, with Exchange alone

This follows Microsoft’s procedure Create a Send connector to route outbound mail through a smart host (Exchange 2016, 2019 and Subscription Edition). The archived article Configure a Dedicated Send Connector for a Specific Domain describes the same idea for Exchange 2010.

1. Keep the default route. You need one send connector with the address space * for all other external mail, either with DNS delivery or with your usual smart host. Most servers already have it.
2. Create the dedicated connector. In the Exchange admin center go to Mail flow > Send connectors > Add. Give it a descriptive name, choose the type Custom, select Route mail through smart hosts and add the smart host by IP address or FQDN. Choose the authentication the smart host requires; if it is a login, use Offer basic authentication only after starting TLS rather than plain Basic authentication, which sends the password in clear text.
3. Set the address space. Add the recipient domain as an SMTP address space, for example partner.example, or *.partner.example to include subdomains, with cost 1. Leave Scoped send connector unticked unless you deliberately want the connector to be visible only inside one Active Directory site. Select the source server and finish.
4. Repeat for every destination that needs its own smart host. Each dedicated connector must carry a different recipient domain.

The same in the Exchange Management Shell, using Microsoft’s documented syntax:

New-SendConnector -Name "Partner via gateway" -AddressSpaces "SMTP:partner.example;1" -Custom -DNSRoutingEnabled $false -SmartHosts gateway.partner.example -SmartHostAuthMechanism None

New-SendConnector -Name "Smart host to Internet" -AddressSpaces * -Custom -DNSRoutingEnabled $false -SmartHosts 192.168.3.2 -SmartHostAuthMechanism None

Verify what Exchange will choose from:

Get-SendConnector | Format-List Name,Enabled,AddressSpaces,SmartHosts,Port,IsScopedConnector,SourceTransportServers

Then send a message to the partner domain and one to any other domain. Microsoft’s own check is the same: send a test message, and turn on protocol logging for the send connector if you want to see the session. Because the more precise address space wins regardless of cost, the partner connector takes mail for partner.example even though the * connector has the same cost of 1.

Two details that save time. A send connector may list several smart hosts, but they are alternatives for one route and share one authentication setting, so this is a redundancy feature, not a way to use two providers. And two connectors with the address space * do not split your mail; the one with the lower cost is preferred for all of it.

Method B: one smart host per sender, with a relay behind a single connector

When the smart host must depend on the sender, Exchange keeps exactly one send connector and hands everything to a local relay that knows the senders. MultiSendcon is that relay: a Windows service that accepts mail from Exchange on an alternative port (Exchange sees it as an ordinary smart host) and distributes it to the SMTP account that belongs to the sender. The full walk-through with screenshots is the MultiSendcon installation guide; in short:

1. Install. Run the installer from the download page on the Exchange server. It starts the service and creates a send connector named MultiSendcon in Exchange with the address space *, port 2500 and a cost of 100, so all mail still takes your existing connectors until you are ready. If the installer cannot create the connector, one command does it (knowledge base: create the MultiSendcon send connector manually):

   new-sendconnector -name MultiSendcon -Addressspaces "SMTP:*;100" -port 2500 -SmarthostAuthMechanism none -smarthosts 127.0.0.1

2. Master relay. In MultiSendcon Administrator, Configure > Master relay, select the local IP address the service listens on and leave the port at 2500. Limit the IP addresses that may connect, so the relay can never be used from outside.
3. One SMTP account per provider. On the Relay servers tab choose Add Relay Server for each smart host: server name, port (typically 25, 587 or 465; SSL is switched on automatically on 465), user name and password, and the sender address filter that decides who uses it, *@company-a.example for a whole domain or user@company-b.example for one address. A blank filter matches every sender. When several accounts match, the topmost entry wins, so put single addresses above their domain; two accounts with the same mask are tried in list order, which gives you a primary and a fallback relay. Add Direct DNS creates an entry that delivers straight to the recipient’s mail server by MX lookup, useful as a fallback for addresses that no relay account matches.
4. NDR handling. Enter the name or IP address of your Exchange server as the SMTP server, so non-delivery reports for mail a provider refuses find their way back to the sender or to a postmaster.
5. Test each account with the built-in SMTP account test. Use a From address that matches the account’s sender filter, and watch the Administrator log for the provider’s answer.
6. Switch Exchange over. On the MultiSendcon send connector change the address space cost from 100 to 1 (enable the connector if Exchange shows it as disabled) and replace the smart host 127.0.0.1 with the IP address you selected on the Master relay tab. In the shell that is Set-SendConnector MultiSendcon -AddressSpaces "SMTP:*;1" -SmartHosts <IP address>. Send a message from each sender domain and follow it in the Administrator log to the right account.

Method A and method B side by side

	A: per recipient domain (Exchange only)	B: per sender (Exchange + relay)
What selects the smart host	The recipient’s domain, through the connector’s address space	The sender’s address or domain, through the account’s sender mask; optionally the recipient as well
Send connectors in Exchange	One per destination plus the * connector	One (*, port 2500, cost 1)
Credentials	One authentication setting per connector	One login per SMTP account, as many accounts as you need
Order of evaluation	Most specific address space, then cost, hop count, connector name	List order in the Relay servers tab; topmost match wins
Fallback	A second connector with the same address space and a higher cost	A second account with the same sender mask below the first
Where you test	Test message plus the send connector protocol log	Account test dialog plus the MultiSendcon Administrator log
Extra software	None	One Windows service on the Exchange server
Typical use	Partner gateway, outbound filtering appliance for some destinations	Several companies or domains on one server, each with its own provider relay

The two methods combine without conflict. A dedicated connector for partner.example keeps winning for that destination, because its address space is more specific than the * of the MultiSendcon connector; everything else goes to the relay and is sorted by sender there.

Frequently asked questions

More on outbound mail flow in the Exchange mail flow guides, or in the knowledge base. Auf Deutsch: Mehrere Smarthosts in Exchange 2016/2019.