Email Verifier and Pattern Finder — No SMTP

Check email syntax, DNS mail routing, provider identity, disposable domains, role accounts, and optional Gravatar presence. Generate ranked address candidates from a person's name and a domain. This actor makes no SMTP connections, sends no email, and never confirms mailbox existence. A deliverable verdict means only that the supported syntax is valid and DNS advertises mail routing, without the risk flags described below. It is not a guarantee that a message will arrive.

Quick start

Use Python 3.12. Create a virtual environment, install the pinned dependencies, and run the actor locally:

python -m venv .venv
.venv/Scripts/python.exe -m pip install -r requirements.txt
apify validate-schema .actor/input_schema.json
.venv/Scripts/python.exe -m unittest discover -s tests -v
./validation/run_live.ps1

The validation script copies INPUT.json into fresh actor-local storage and runs the actual Apify SDK entry point. It requires no keys, forces local execution in the child process, and exports counts, timings, and rows under validation/. Pass -Python path/to/python.exe to use an existing compatible interpreter. On Linux, use .venv/bin/python validation/live.py. For ordinary execution, supply input through Apify's default key-value store and run python -m src or apify run.

The supplied daily sample contains 30 mixed addresses and three finder entries. It includes public role addresses, synthetic test identifiers, obvious syntax failures, disposable services, typo domains, and one blank. Names and known examples in finder samples are illustrative; they are not evidence of actual people or contacts at those organizations.

Input

emails accepts up to 50,000 strings. Whitespace around an address is trimmed; domain names are lowercased and converted to IDNA ASCII. Local-part case is preserved. The practical RFC-style subset supports dot-atoms, plus tags, and apostrophes. Quoted local parts, comments, domain literals, and internationalized local parts requiring SMTPUTF8 are unsupported and marked invalid. Invalid syntax is still a completed, billable address check.

finder accepts up to 5,000 objects containing firstName, lastName, and domain. Each may include up to 1,000 knownExamples: email strings or objects with email, firstName, and lastName. Named examples provide exact pattern evidence. Bare addresses provide weak separator evidence; ambiguous unseparated strings do not identify a pattern. Other-domain and invalid examples are ignored.

checkGravatar defaults to true. Set it false to avoid sending address hashes to Gravatar. concurrency defaults to 20 and separately caps DNS lookups, HTTP requests, and input batch size. timeoutSecs defaults to five seconds per network attempt. HTTP 429/5xx responses and transport failures receive two retries, with one- and two-second backoff. Transient DNS failures receive the same bounded retry schedule. The timeout is not a whole-run deadline; large lists naturally take longer.

Checks and interpretation

Address rows include email, normalized, valid_syntax, suggestion, domain, domainExists, hasMx, mxHosts, mxProvider, disposable, role, freeProvider, catchAllLikely, gravatar, verdict, reasons, and source. Additional rowType identifies verify, finder, error, or summary rows.

domainExists specifically reports A/AAAA presence; an MX-only domain can legitimately have false here and still accept mail. DNS uncertainty produces null rather than false. MX provider identification uses hostname suffixes and may identify a filtering gateway rather than the ultimate mailbox host. Null MX explicitly rejects mail. No MX and no address records is undeliverable; address-only fallback is risky.

Disposable, role, suspected typo, and catch-all signals make otherwise plausible addresses risky. Free-provider membership alone does not. Suggestions are never silently applied. Catch-all likelihood reflects configurable wildcard/disposable providers or address-only fallback; it does not establish that arbitrary recipients are accepted. Unknown DNS results remain risky, not proof of rejection. Gravatar returns true for HTTP 200, false for 404, and null when disabled, skipped, or unavailable. Avatar presence never upgrades a verdict.

Finder results and pricing

Finder rows contain names, domain, candidates, best, source, and an explicit qualification. Ten patterns include first.last, first, flast, first_last, f.last, firstl, last.first, firstlast, last, and lastf. Names are transliterated to ASCII letters; unusable names produce free errors. Duplicate candidates are removed. Each candidate contains its email, pattern, verdict, ranking score, reasons, and full checks. Scores are ordering weights, not probabilities. best is null when every candidate is undeliverable. These are guesses: “syntactically plausible + domain accepts mail” always carries the DNS-only qualification.

Pricing is $0.0005 per processed address and $0.004 per completed finder entry, including all candidate checks. Blank/malformed input errors and empty-input summaries are uncharged. One Actor.charge call precedes each billable row; exhausted budgets stop output. Charging and persistence are not atomic. Configure the two custom events and disable synthetic events before publication; local runs do not bill.

The bundled disposable-email-domains list includes its CC0 license, pinned source commit, and checksums in vendor/. Subdomains inherit list matches. Lists and provider mappings can become stale; review updates periodically. See VALIDATION.md for observed results and deployment limits.

Example output

One recorded dataset row, trimmed by omitting fields only. Values are the saved snapshot, not current measurements. Source: validation/rows.json, first row.

{
  "rowType": "verify",
  "normalized": "info@python.org",
  "valid_syntax": true,
  "hasMx": true,
  "role": true,
  "verdict": "risky",
  "reasons": [
    "Syntactically plausible + domain accepts mail according to DNS only; mailbox unconfirmed",
    "Shared or automated role account",
    "Mailbox existence unconfirmed; no SMTP used"
  ]
}

Use cases

• A CRM administrator can flag syntax errors and suspected domain typos before importing contacts, keeping suggested corrections for human review.
• A newsletter operations team can separate disposable and role addresses for its own list-quality policy without treating DNS as mailbox confirmation.
• A recruiting researcher can generate candidate patterns from a supplied name and domain, using known examples to rank guesses for independent verification.
• A customer support team can inspect mail-routing and provider signals when investigating address problems, preserving unknown DNS outcomes instead of assuming rejection.

Pricing example

Hypothetical batch, calculated from .actor/pay_per_event.json:

Event	Count	USD per event	Subtotal
email-verified	10,000	$0.0005	$5.0000
finder-query	100	$0.004	$0.4000

Total declared event charges: $5.40. These counts are a budgeting example, not a promised yield or an actual bill. Any applicable platform or proxy costs are outside this calculation.

Processed invalid-syntax addresses count as checks. Finder candidates are included in the finder event, with no separate per-candidate charge.

Limitations

No SMTP or mailbox-existence check is performed. Finder scores rank guesses rather than estimate delivery probability. The saved public role address is classified risky; it is not a verified contact.