Stop Bounces Before You Ever Send.

Every address runs a full RFC 5321 and RFC 5322 compliance pass before a single network call goes out. The engine catches what visual scanning misses, the double dot in [email protected], the trailing period, the IDN homograph that looks valid but resolves to a different domain.

Bundled typo suggestions let your form offer “did you mean [email protected]?” instead of rejecting silently.

Once syntax passes, the engine resolves the domain. We confirm the DNS records exist, fetch the MX record priority list in order, and verify at least one mail-exchange server is actively accepting connections right now.

Misspelled domains like gmial.com, expired domains, and parked-for-sale domains all fail this gate before the engine wastes a single SMTP roundtrip.

The engine opens a TCP connection on port 25, performs the EHLO handshake, then negotiates MAIL FROM and RCPT TO. Every server response code (220, 250, 550, 552) is parsed deterministically against the IANA enhanced-status registry.

This is the moment a mailbox proves it actually exists. No third-party guesses, no statistical heuristics, just the receiving server's own answer.

Some domains accept every email regardless of whether the mailbox exists, a setup known as a catch-all configuration. The engine sends a deterministic probe to a deliberately fake address ([email protected]); if the server returns the same 250 OK it returned for the real address, the domain is catch-all.

The verdict isn't dropped, it's flagged RISKY so you know the deliverability signal is degraded.

The engine maintains a curated registry of 10,247 disposable email providers, including Mailinator, Guerrilla Mail, 10MinuteMail, Tempmail, and the long tail of regional clones.

Any address matching the blocklist is flagged INVALID. Deliverability to a mailbox that exists for 10 minutes and is never checked is functionally zero, regardless of whether the SMTP handshake passes.

info@, support@, no-reply@, admin@, billing@. These are shared inboxes, not individuals.

The engine extracts the local-part of every address, matches it against the known role-prefix registry, and tags the result with a reduced engagement score.

You don't drop them automatically. The verdict flags them as roles so you can decide.

Fresh-spam domains registered hours ago are the single biggest source of inbound abuse. The engine queries WHOIS and RDAP for every unique domain, extracts the registration date, and flags anything under 30 days old with a “fresh” warning.

Domains aged 5+ years pick up a corresponding trust signal. The same heuristic spam filters have been using since the early 2000s, ported into the verdict.

SPF, DKIM, and DMARC together prove the sender is authorised to send from that domain.

The engine reads each policy via DNS, validates SPF includes recursively, scans six common DKIM selectors, and confirms DMARC alignment with the From: header.

A failing DMARC policy means the sender can be spoofed, so the verdict warns you before you reply.

Beyond “exists vs doesn't exist”, the engine extracts the precise mailbox state from the SMTP server's response. Full inbox (552 / 522 quota), disabled mailbox (550 5.1.1), out-of-office autoresponder, frozen account.

Each state maps to a specific retry policy. Full inbox retries in 6 hours. Disabled drops permanently. The verdict tells you which bucket the bounce belongs in so your retry logic doesn't waste cycles.