uCheckeruChecker

SMTP Test

Connect to a mail server the way another mail server would, and see what it says: the greeting, whether it offers STARTTLS, what certificate it presents, whether its reverse DNS lines up, and whether it will carry mail that is not its own.

Beta

This tool has just launched. Unlimited checks, no sign-up, no payment. If a result looks wrong, tell us — that is the most useful feedback we can get right now.

What this actually does

It opens the same conversation any other mail server would open. Connect, read the greeting, say hello, look at what the server advertises, and, if it offers encryption, negotiate it and inspect the certificate.

No message is ever sent. The exchange stops before the command that would carry one, and when a relay probe is run it is followed by a reset rather than any content. What you see is what a receiving server sees in the first second of a delivery attempt.

The three names that have to agree

This is the finding that surprises people most, because it has nothing to do with the records they spent the afternoon configuring.

When a server connects, the receiver knows one thing for certain: the address it came from. It looks that address up in reverse DNS to get a name, then resolves that name forward to see whether it points back. It also compares both against the name the sender announced in HELO. Three values, and they are expected to describe one machine.

  • No PTR at all. A large share of receivers reject the connection immediately, before any authentication is considered.
  • A PTR that does not resolve back. Anyone can point a name at an address; only the address owner can make the reverse record. When the forward lookup does not return the same address, the name proves nothing.
  • HELO disagreeing with the PTR. The usual result of a default configuration announcing an internal hostname such as localhost.localdomain or a container ID.

Getting this right is not subtle work: one PTR record from whoever owns the address, one hostname in the mail server's configuration, and an A record that ties them together. It is simply invisible from the DNS side, which is why it survives so many otherwise careful setups.

Encryption, and what the certificate has to do

Mail between servers uses opportunistic TLS: the sender asks whether encryption is available and uses it if so. The certificate is usually not validated, which is why self-signed certificates carry mail every day without complaint.

Two things change that. Publishing an MTA-STS policy tells senders to require TLS with a valid, matching certificate and to refuse delivery rather than fall back to plaintext, so a certificate that was fine yesterday starts bouncing mail. And since February 2024 Gmail requires TLS from every sender delivering to its mailboxes.

The expiry date is worth a note in the calendar for the same reason. A certificate that lapses on an ordinary web server produces a warning someone can click through. On a mail server under MTA-STS it produces silence, and mail that stops arriving.

Open relay: rare, and catastrophic

A relay that carries mail for anyone will be found within days of appearing, and the address will be on every blocklist shortly after. It is rare in modern software because no current mail server ships that way. It turns up after a hand-edited configuration, or a permissive rule added to solve a delivery problem and never removed.

When you test a domain, we ask its mail server to accept a message from an address it has no relationship with, addressed to a domain it does not host, and stop before sending anything. A correct server refuses. A server that agrees is being abused already, whether or not anyone has noticed.

We do not run this probe against a specific host you type in. Asking an arbitrary machine whether it will carry someone else's mail is indistinguishable from scanning, and our own address would be the one paying for it.

Closing an open relay

Open relays used to be normal: the first SMTP specification, RFC 821 from 1982, had no authentication at all. That changed once spammers started using other people's servers at scale, and by the late 1990s an open relay was treated as a vulnerability.

To test by hand, connect to port 25 and send three commands:

EHLO test.example.com
MAIL FROM:<test@test.example.com>
RCPT TO:<someone@gmail.com>

If RCPT TO for an outside address is answered with 250 before any authentication, the relay is open. A correct server replies 550 or 554 with something like Relay access denied.

In Postfix the fix lives in smtpd_relay_restrictions. A typical setting lets through only your own network and authenticated users:

smtpd_relay_restrictions =
    permit_mynetworks,
    permit_sasl_authenticated,
    reject_unauth_destination

An open relay also breaks SMTP address verification, because it answers 250 for every mailbox and a nonexistent address looks deliverable. A validator recognises such servers and does not trust their replies.

Reading the capability list

The server lists what it supports in response to the greeting. A few of those lines are worth knowing.

  • STARTTLS. Encryption is available. Its absence on a public mail server is a real gap in 2026.
  • SIZE. The largest message the server will accept. Worth checking before blaming a bounce on filtering.
  • AUTH. Authenticated submission is offered. Expected on port 587, and a little unusual on port 25.
  • 8BITMIME and SMTPUTF8. The server handles non-ASCII content and international addresses without re-encoding.

The list often changes after encryption is negotiated: many servers advertise authentication only once the channel is secure. We re-issue the greeting after the upgrade and show what the server says then, since that is what a real sender would act on.

What an SMTP relay is

A relay is any mail server that passes a message on without being its sender or its final recipient. The word covers three different things. There is relaying as the basic mechanism, which is how all server-to-server delivery works. There is an SMTP relay service, a hosted server your application hands mail to so you do not have to run and warm up your own. And there is the open relay, a server that forwards anything for anyone. The first is the protocol, the second is normal infrastructure, the third is an incident.

With a relay service, your application connects on port 587 with STARTTLS or on 465 with implicit TLS, authenticates, and hands the message over. The service then looks up the recipient's MX, negotiates TLS with that server, retries on 4xx replies and turns 5xx replies into bounce events. On your side it needs three DNS changes: the service's include in SPF, a DKIM key (usually a CNAME to the service's key, or your own key), and a custom Return-Path on a subdomain of yours, so that SPF aligns with the From domain and DMARC passes. For most teams a paid relay is a saner choice than running a mail server of their own; see the comparison of SMTP relay services.

Running your own relay server

Companies still run their own relays for down-to-earth reasons. Many hosting providers block outbound port 25, so a server there can only send through an authenticated relay. A company with hundreds of internal servers would rather push all outgoing mail through one gateway that holds the queue, signs with DKIM, scans for malware and keeps a single log. And if one internal box is compromised, you cut it off at the relay instead of letting it burn the reputation of the whole domain.

In Postfix and Sendmail this is called a smart host. Postfix needs one line, and from then on it stops doing its own MX lookups and passes everything to the relay:

relayhost = [smtp-relay.example.com]:587

On a shared relay IP your reputation depends on the neighbours; a dedicated IP gives you control but has to be warmed up. Every relay on the path adds its own Received header, and read bottom to top they give the route of the message with timestamps. The email header analyzer lays that route out for you.

SMTP reply codes

Every command gets a three-digit reply. The first digit carries the meaning: 2 is success, 3 means the server is waiting for more, 4 is a temporary failure and 5 is a permanent one. The difference between 4 and 5 matters more than the exact code, because the sender retries a 4xx on its own and never retries a 5xx. The base codes come from RFC 5321, while 235 and 334 were added by the AUTH extension in RFC 4954. The second digit names the category: 0 syntax, 1 information, 2 connection, 5 mail system.

  • 220, 221. The server is ready; the server is closing the connection after QUIT.
  • 250, 251. The command was completed; 251 means the recipient is not local and the server will forward the message.
  • 252. A VRFY reply: the server cannot confirm the mailbox but will accept mail for it.
  • 235, 334. Authentication succeeded; the server is waiting for the next AUTH step.
  • 354. Ready for the message body after DATA.
  • 421. Service not available and the connection is being closed: overload, maintenance or rate limiting.
  • 450, 451, 452. Temporary failures. Greylisting usually shows up as 450 or 451; 452 means insufficient storage or too many recipients in one transaction (RFC 5321, section 4.5.3.1.10).
  • 455. The server cannot accommodate the command parameters, usually an ESMTP extension issue.
  • 550. Either the mailbox does not exist (hard bounce) or the sender was refused on policy grounds. The text after the code tells which.
  • 551, 553. The recipient is not on this server and it will not forward; the mailbox name is not allowed, usually a syntax error.
  • 552. Storage exceeded, for example a message larger than the SIZE the server advertises (RFC 1870).
  • 554. The transaction failed, often with a note about reputation or a blocklist.

Next to the code there is often an enhanced status code of three numbers separated by dots, and after it a line of text written by the server. Read the text: the same code is used for different reasons by different servers.

  • 5.1.1 the mailbox does not exist; 5.1.2 the destination domain does not exist or takes no mail.
  • 4.2.1 / 5.2.1 the mailbox is disabled.
  • 5.2.2 mailbox full. It is a class 5 code, but RFC 3463 suggests treating a full mailbox as a persistent transient condition, since the owner can free up space.
  • 5.7.1 refused on policy grounds; 4.7.0 / 4.7.1 temporary refusals from the security and policy class. RFC 3463 defines X.7.0 as other security status and X.7.1 as delivery not authorized; the text after the code names the actual reason.
  • 5.7.26 multiple authentication checks failed (RFC 7372); Gmail also returns it when DMARC fails.

Reply codes in address verification

The uChecker validator asks a server the same RCPT TO question a sender would, and never gets as far as DATA. A 250 means the mailbox exists. A 550 with user unknown or 5.1.1 makes the address bad, while a 550 about a blocklist or policy is about the sender, so we re-check that address from another IP. A 450 or any other 4xx is simply retried later. Before the real addresses, the validator asks about one that cannot exist: if the server accepts that too, the domain is catch-all and a 250 for the rest proves nothing. Servers that accept everything at RCPT TO and only refuse after DATA land in the same bucket.

Sources

SMTP questions

Read next

Server configured correctly and mail still bounces?

A perfect handshake does not help when the recipient address stopped existing months ago. uChecker verifies your list before you send, and the first 100 addresses are free.

Check your list