Signet. Docs

API documentation

One call. A verdict you can route on.

Signet scores an email address, phone number or IP address for signup and order fraud. Each call returns a verdict (allow, review or block), a risk score from 0 to 1, and the named signals behind it.

Every check runs through two layers:

  1. Free rules. Syntax, typo domains, role and throwaway patterns, a disposable-domain registry synced daily, reserved and Tor IP ranges, phone numbering plans. A check rejected here costs nothing.
  2. Deep check. Live DNS deliverability plus Signet's network reputation: fraud reports and flags from other merchants, and how often the same value was checked in the last 24 hours. A check that reaches this layer costs 1 credit.

Base URL for every endpoint:

https://puresignet.com/v1

Quickstart

Send the value you want to check. The type is detected automatically.

curl https://puresignet.com/v1/validate \
  -H "X-API-Key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"input": "jane@company.com"}'

The response:

{
  "input": "jane@company.com",
  "type": "email",
  "normalized": "jane@company.com",
  "valid": true,
  "verdict": "allow",
  "risk_score": 0,
  "confidence": 0.88,
  "fraud_indicators": [],
  "free_provider": false,
  "network": {
    "merchants_seen": 0,
    "merchants_flagged": 0,
    "fraud_reports": 0,
    "velocity_24h": 0
  },
  "layer": "deep_network",
  "credits_used": 1,
  "cached": false,
  "response_time_ms": 142,
  "credits_remaining": 99
}

Prefer clicking to typing? The playground runs the same calls with your key.

Authentication

Send your API key in the X-API-Key header on every request. Keys look like ak_live_ followed by 32 hexadecimal characters.

  • Create keys in your dashboard, which you sign in to with a link sent to your email. A new key is shown once; Signet stores only a hash of it and cannot show it again.
  • Your plan sets how many keys you can have at once: 1 on the trial and Starter, 3 on Professional, 10 on Scale. Use separate keys for production, staging and each app, so one can be cut off without touching the others. All of an account's keys share its balance and its per-minute limit.
  • Call the API from your server. Never put the key in browser or mobile app code, where anyone can read it.
  • If a key leaks, replace it in the dashboard: the old key stops working at once and a new one takes its place. Revoke a key you no longer use.

Validate

POST/v1/validate0 or 1 credit

Scores one email address, phone number or IP address.

Request body

FieldTypeRequiredDescription
inputstringyesThe value to check. Up to 320 characters.
typestringnoemail, phone or ip. Leave it out to auto-detect. Send it when you know it, so a mistyped email is judged as an email.
countrystringnoTwo-letter country code such as IN. Used to read phone numbers written without a +, e.g. 9820034571.
livebooleannoAsk for a live check. See Live checks.

Phone numbers work best in international format with a leading + and country code, for example +919820034571. If your customers type local numbers, send country with them. IPv4 and IPv6 are both supported.

What you get back for each type

TypeEnrichment
EmailDeliverability of the domain, disposable and typo detection with a suggested fix, whether it is a consumer mailbox.
PhoneCountry, line type (mobile, landline, voip, toll_free, premium_rate), international format, and whether the number sits in a range the country has issued.
IPThe network that owns the address, its AS number, where that network is registered, and whether it is a hosting provider rather than a home or mobile connection.

The response is described under Response fields.

Live checks

A normal check works from rules, DNS and network reputation. A live check goes one step further and asks the source directly. Send "live": true:

TypeWhat is askedYou getExtra cost
EmailThe mail server: does this exact mailbox exist?mailbox: deliverable, undeliverable or catch_all+1 credit
PhoneThe mobile network: is this number in service right now?line_status: connected, absent or not_in_service, plus carrier and ported+2 credits
curl https://puresignet.com/v1/validate \
  -H "X-API-Key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"input": "+919820034571", "live": true}'
  • The extra credits are charged only when the source answers. Each result carries "live": true if it did. If it could not be reached, you get the normal verdict at the normal 1 credit and "live": false.
  • A live check is skipped, and costs nothing extra, when the free rules or DNS already settled the answer.
  • Live checks take longer, sometimes a few seconds. Use them where the decision matters: before shipping a cash-on-delivery order, before paying out, before sending an expensive campaign.
  • Availability. Live checks are being switched on gradually. Until they are on for your account, "live": true is accepted and simply returns "live": false.

VPN, proxy and Tor detection for IP addresses is part of every normal check and costs nothing extra.

Validate batch

POST/v1/validate/batch0 or 1 credit per unique value

Scores up to 100 values in one call.

Request body

FieldTypeRequiredDescription
itemsstring[]yes1 to 100 values, each up to 320 characters.
typestringnoApplies one type to every item. Leave it out to auto-detect each one.
countrystringnoTwo-letter country code used for phone numbers written without a +.
curl https://puresignet.com/v1/validate/batch \
  -H "X-API-Key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"items": ["jane@company.com", "x@mailinator.com", "8.8.8.8"]}'

Response

{
  "results": [ { "input": "jane@company.com", "verdict": "allow", "credits_used": 1, "…": "…" } ],
  "summary": {
    "total": 3, "valid": 2, "invalid": 1,
    "allow": 2, "review": 0, "block": 1,
    "credits_used": 2
  },
  "credits_remaining": 97
}
  • results is in the same order as items. Each entry has the same shape as a single validate response.
  • A value that appears more than once is charged once. Repeats come back with "cached": true and "credits_used": 0.
  • A batch is all-or-nothing. If your balance cannot cover it you get a 402 and nothing is charged.

Feedback

POST/v1/feedbackfree

Tell Signet how a signup or order actually turned out. Report a chargeback, a fake account or a returned cash-on-delivery order as fraud; report a confirmed good customer as legit. The next time anyone checks the same value, the verdict reflects it.

FieldTypeRequiredDescription
inputstringyesThe email, phone number or IP the outcome is about.
outcomestringyesfraud or legit.
typestringnoSame as on validate.
curl https://puresignet.com/v1/feedback \
  -H "X-API-Key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"input": "fraudster@example.com", "outcome": "fraud"}'

Each account counts as one vote per value, and sending a new outcome replaces your earlier one. A single report moves a value towards review; it takes reports from several independent merchants to block it.

Credits

GET/v1/creditsfree
curl https://puresignet.com/v1/credits -H "X-API-Key: YOUR_API_KEY"
{ "balance": 97, "api_key": "ak_live_9f3c…", "message": "Use /usage to see your usage history" }

You rarely need this: every validate response already carries credits_remaining.

Usage

GET/v1/usage?limit=50free

Totals for your key and your most recent checks. limit is 0 to 1000 (default 50); use 0 for totals only.

{
  "email": "you@company.com",
  "balance": 97,
  "total_validations": 3,
  "total_credits_used": 2,
  "recent_usage": [
    {
      "id": 1042, "input": "j***@company.com", "type": "email",
      "valid": true, "credits": 1, "layer": "deep_network",
      "risk_score": 0, "created_at": "2026-11-14T05:31:09Z"
    }
  ]
}

Inputs in usage history are masked. See Data handling.

Test your key

POST/v1/testfree

Confirms a key works and returns the account it belongs to. Use it as a health check when you deploy.

curl -X POST https://puresignet.com/v1/test -H "X-API-Key: YOUR_API_KEY"

Response fields

FieldTypeDescription
verdictstringallow, review or block. The field to act on. See Verdicts.
risk_scorenumber0 (no risk found) to 1 (certain). Combines every signal.
validbooleantrue when the value is well-formed and the risk score is under 0.5.
confidencenumber0 to 1. How decisive the evidence is.
fraud_indicatorsstring[]Names of the signals that fired. Empty when nothing did. See Fraud signals.
networkobject | nullCounts from other merchants: merchants_seen, merchants_flagged, fraud_reports, and velocity_24h (checks of this value in the last 24 hours). null when the input was malformed.
typestringemail, phone or ip.
inputstringWhat you sent.
normalizedstringCleaned form: lower-cased email, digits-only phone with +, trimmed IP.
layerstringWhich layer settled it. Starts with local_ for the free rules; deep_network for the deep check.
credits_usedinteger0 or 1.
credits_remainingintegerYour balance after this call.
cachedbooleantrue for a repeated value inside a batch.
response_time_msintegerServer processing time.
suggestionstringEmails only, when present: the likely intended address, e.g. john@gmail.com for john@gmial.com.
free_providerbooleanEmails only: a consumer mailbox such as Gmail or Outlook.
countrystringPhones only: two-letter country code, e.g. IN.
line_typestringPhones only: mobile, landline, landline_or_mobile (countries such as the US that do not separate them), voip, toll_free, premium_rate or other.
formattedstringPhones only: the number in international format, e.g. +91 98200 34571.
providerstringIPs only: the organisation that operates the network.
asnintegerIPs only: the network's AS number.
proxy_typestringIPs only, when detected: vpn, tor or proxy.
livebooleantrue when a live check answered for this item.
domain_age_daysintegerEmails on a business domain, when known: days since the domain was registered (from public registry data). Not given for free mailbox providers such as Gmail.
mailboxstringEmails, live check only: deliverable, undeliverable or catch_all (the server accepts any name, so the inbox cannot be confirmed).
line_statusstringPhones, live check only: connected, absent (switched off or out of coverage) or not_in_service.
carrier, portedstring, booleanPhones, live check only: the network the number is on now, and whether it moved there from another.
network_countrystringIPs only: where the network block is registered. A strong hint about location, not a geolocation of the person.

Verdicts

VerdictWhenSuggested action
allowRisk score under 0.3 and nothing invalid.Let the signup or order through.
reviewRisk score from 0.3 up to 0.7, or the value failed validation without being clearly fraudulent.Add friction: email or OTP confirmation, prepaid instead of cash on delivery, or a manual look. Signet does not hold or review anything itself: your code decides, and treating review as allow is a valid choice.
blockRisk score 0.7 or higher, or the value is malformed.Reject, and ask the person for a different email or number.

If your tolerance differs, ignore the verdict and set your own thresholds on risk_score.

Credits & billing

What happenedCredits
Rejected by the free rules (malformed, disposable, reserved range, obvious fake pattern)0
Passed the free rules and went through the deep check, whatever the final verdict1
Live mailbox check that answered ("live": true on an email)+1
Live number check that answered ("live": true on a phone)+2
Live check requested but the source did not answer+0
Repeated value inside one batch0
Feedback, credits, usage and test calls0
Any request that returns an error0

Trial and plans. Every account starts with a 7-day free trial that includes 100 deep checks; unused trial checks end with the trial. Each paid plan adds its monthly credits at the start of every billing period. On Professional unused credits stay usable for one extra month, on Scale for three; on Starter they end with the month. The oldest credits are always spent first.

Email, phone and IP checks all cost the same. When the balance reaches zero, deep checks return 402 insufficient_credits; values the free rules reject are still answered.

Fraud signals

These are the names that appear in fraud_indicators. Names ending in _invalid mean the value is malformed.

Network (any type)

reported_fraud_networkOther merchants reported this value as fraud.
flagged_across_networkThe value was judged risky at two or more other merchants.
high_velocityChecked unusually often in the last 24 hours.

Email

disposable_domain, disposable_patternA throwaway inbox provider.
no_mx_recordsThe domain cannot receive email.
new_domainThe email's domain was registered in the last 30 days (stronger under 7). Real businesses rarely sign up from a domain that is days old. domain_age_days gives the age.
mailbox_not_foundLive check: the mail server says this mailbox does not exist.
catch_all_domainLive check: the server accepts mail for any name. Mild on its own.
provider_typoLooks like a mistyped major provider. A suggestion is included.
provider_spoofingA major provider's name embedded in another domain, e.g. gmail.com.example.tk.
spam_trap_domainA known spam-trap domain.
fake_pattern, keyboard_walk, gibberish_local_partThe part before the @ looks made up: test, asdfgh, random characters.
role_based_addressA shared mailbox such as admin@ or sales@.
plus_alias, numeric_local_part, numeric_domainWeak signals that add up with others.
suspicious_tld_…A top-level domain heavily used for abuse; the suffix names it.
unicode_confusable_domain, unicode_confusable_localLook-alike characters from another alphabet.
free_email_providerInformational. Carries almost no risk on its own.
syntax_invalidlength_invaliddouble_dot_invalidedge_special_invaliddomain_hyphen_invalid

Phone

number_not_in_serviceLive check: the mobile network says this number is not assigned to anyone.
handset_unreachableLive check: a real number, but the phone is off or out of coverage right now. Mild on its own.
voip_numberAn internet phone number: cheap to create in bulk and often unable to receive OTPs.
non_personal_lineToll-free, premium-rate or another kind of line that is not a person's own phone.
landlineA fixed line. Mild on its own, but it cannot receive an SMS.
unassigned_number_rangeThe right length for its country, but not in a range the country has issued.
all_same_digit, sequential_digits, repeating_pairs, long_repeat, palindrome, repeating_last4Digit patterns people type when inventing a number.
reserved_555_range, tollfree_555_combo, uk_fictional_rangeRanges reserved for fiction and testing.
premium_rate_prefix, n11_service_codePremium-rate or service numbers, not personal lines.
invalid_area_code, unassigned_area_code, invalid_exchange_codeNot a number the North American plan can assign.
length_for_country_mismatch, invalid_country_code, leading_zerosWrong shape for the country code given.
letters_invalidlength_invalidnumber_length_invalid

IP address

tor_exit_nodeA known Tor exit range.
vpn_detected, proxy_detected, tor_networkThe address is a known VPN server, open proxy or Tor exit. proxy_type says which.
hijacked_networkThe address is in a network Spamhaus lists as hijacked or run by criminals (the DROP list). Block.
hosting_networkThe address belongs to a cloud or hosting provider (including the ranges AWS, Google Cloud, Oracle and DigitalOcean publish). Real customers rarely browse from a server; bots, scrapers and VPN users do.
datacenter_range, cdn_rangeA hosting or CDN network, not a home or mobile connection. provider names it.
private_rfc1918, loopback, link_local_apipa, carrier_grade_nat, unique_local, link_localNot a public address. Usually means your server is passing its own or a proxy's IP instead of the visitor's.
documentation_test_net, benchmarking, multicast, broadcast, reserved_future_use, this_network, unspecified_addressReserved ranges that never belong to a real visitor.
teredo_tunnel, 6to4_tunnel, nat64_well_known, discard_onlyIPv6 tunnelling and special-purpose prefixes.
format_invalid

Errors

Errors use standard HTTP status codes and always return JSON with a human-readable error and a stable code to branch on:

{ "error": "Not enough credits for this request.", "code": "insufficient_credits", "needed": 3, "balance": 1 }
StatusCodeMeaning
400bad_input, bad_type, bad_items, too_many_items, bad_country, bad_live, bad_outcome, bad_jsonSomething in the request body is missing or wrong. The message says what.
401invalid_api_keyThe key is missing, malformed or unknown.
402insufficient_creditsThe balance cannot cover this request. Includes needed and balance. Nothing was charged.
402trial_endedThe 7-day free trial is over. Choose a plan in the dashboard.
402no_planThe account has no active plan.
403account_suspendedThe account is paused. Contact support.
405method_not_allowedWrong HTTP method for this endpoint.
413payload_too_largeThe request body is over 64 KB.
422bad_inputFeedback only: the value is not a well-formed email, phone or IP.
429rate_limitedToo many checks this minute for your plan. Wait for the number of seconds in the Retry-After header.
500internal_errorOur fault. The body includes a request_id; send it to support and we can find the exact request.

Limits

Rate limitPer account, by plan: 300 a minute during the trial, then 60 (Starter), 300 (Professional) or 1,200 (Scale). A batch counts one per item. Over the limit you get 429 with Retry-After.
Batch size100 items.
Input length320 characters per value.
Request body64 KB.

Need more? Email hello@puresignet.com.

Data handling

  • Masked logs. Signet does not store the values you check. Logs keep a masked form only, such as j***@company.com, +919***78 or 203.0.*.*, and are deleted after 90 days.
  • Network reputation. To recognise a value seen at another merchant, Signet keeps a keyed one-way hash of it alongside the outcome. The hash cannot be turned back into the email, phone number or IP, and no merchant can see another merchant's data, only the counts in network.
  • API keys. Stored as hashes. Nobody at Signet can read your key.
  • Transport. HTTPS only.
Because results draw on reports shared between merchants, mention in your own privacy policy that you use a fraud-prevention service that processes contact details and IP addresses.

Data sources

Alongside its own rules and the reports merchants send, Signet uses these free public sources. The lists are refreshed daily; domain ages are looked up when first needed and remembered.

SourceUsed forTerms
disposable-email-domainsThrowaway email domainsCC0 / MIT
FakeFilterThrowaway email domainsBSD-3-Clause, © 2022 7c
Tor Project exit listTor exit nodesPublic
Spamhaus DROPHijacked networksFree, with credit to The Spamhaus Project
AWS, Google Cloud, Oracle Cloud, DigitalOceanCloud and hosting server ranges, as each provider publishes themPublished by each provider
RDAP (domain registries, via IANA)When an email domain was registeredPublic registry data
Team Cymru IP-to-ASN, Google and Cloudflare public DNSNetwork owner and mail-server checksFree public services
libphonenumberPhone country, line type and numbering plansApache 2.0

Hijacked-network data: The Spamhaus Project, spamhaus.org/drop.

Integration tips

  • Check at the moment of decision. Call validate when the signup form or checkout is submitted, on your server, before you create the account or accept the order.
  • Pass the visitor's real IP. Behind a load balancer or CDN, read it from the forwarded-for header. A private_rfc1918 result means you sent your own.
  • Fail open. Set a timeout of a few seconds. If Signet is unreachable or returns a 5xx, let the signup through rather than blocking real customers.
  • Use review. It is where most of the value is: confirm by OTP, or switch cash on delivery to prepaid, instead of a hard yes or no.
  • Show the suggestion. When suggestion is present, ask "Did you mean …?" on the form. It rescues real customers who mistyped.
  • Selling cash on delivery? Treat voip_number, landline and hosting_network as reasons to ask for prepayment or an OTP before you ship.
  • Send feedback. Report chargebacks, fake accounts and returned orders. It is free and it improves your own future verdicts as well as the network's.
  • Back off on 429. Honour Retry-After, and prefer one batch call over many single calls for lists.

Roadmap

What we are working on, in order. Last updated November 2026. Nothing here is a promise of a date; if something on this list is what you are waiting for, tell us and it moves up.

StatusWhat
LiveEmail, phone and IP scoring; batch; fraud feedback and network reputation; phone country and line type; IP network owner, hosting, VPN, proxy and Tor detection.
Rolling outLive checks: mailbox existence for emails, in-service status and current carrier for phones.
NextSign up and get a key instantly, without waiting for approval. Pay by card and top up credits yourself.
LaterWebhooks for verdict changes. A Shopify app for cash-on-delivery risk. A public status page.