beta

API

Every page is built from these endpoints. Send an API key as Authorization: Bearer vpndb_…; create one under Account.

This page

GET /api/v1/meta

curl -s -H 'Authorization: Bearer vpndb_EXAMPLE_NOT_A_REAL_KEY' \
  '/api/v1/meta'

All endpoints

GET /api/v1/ip/{ip}Who operates an address
GET /api/v1/ip/{ip}/historyEvery week of an address
GET /api/v1/prefix/{cidr}Any range, and who is inside it
GET /api/v1/prefix/{cidr}/addressesEvery address of a prefix, in a window
GET /api/v1/prefix/{cidr}/timelineHow a prefix was probed, day by day
GET /api/v1/asn/{asn}A network and its providers
GET /api/v1/asn/{asn}/networksEvery prefix of a network
GET /api/v1/org/{id}A company and all its networks
GET /api/v1/org/{id}/networksEvery prefix of a company
GET /api/v1/country/{cc}The infrastructure located in a country
GET /api/v1/countriesEvery country at a glance
GET /api/v1/service/{tag}A provider's profile
GET /api/v1/service/{tag}/networksEvery network a provider exits from
GET /api/v1/servicesEvery provider at a glance
GET /api/v1/torThe Tor network at a glance
GET /api/v1/tor/relaysEvery running Tor relay, for detection
POST /api/v1/bulkMany addresses at once
GET /api/v1/searchResolve free text
GET /api/v1/suggestSuggestions as you type
GET /api/v1/metaThe dataset
GET /api/v1/openapi.yamlThis specification
GET /api/v1/sourcesWhere the data comes from
API documentationOpenAPI specification
Sign in

Entities

Addresses

A single address: which operator runs it, how it was discovered, and what else sits on its network.

The most common entry point. Shows the verdict, the observations behind it, the surrounding network, and where to go next.

Two discovery channels

An address reaches the dataset by one of two independent channels. The distinction bears on how an answer should be weighted.

ChannelWhat happenedMethod
ProbingWe connected through the provider and observed this exitprobe
ScanningThe host answered on a port VPN infrastructure usesscan

Probing is the stronger claim: traffic was routed through the service and egressed here. Scanning establishes that infrastructure is present without proving egress. Both are evidence; only probing produces a time series.

An address neither channel reached is still answered when any of our sources — probing, the enriched snapshot, scan certificates — knows a VPN exit in its announced prefix. The answer then carries method: "interpolated", names the prefix in network, gives the prefix’s VPN providers and weights, and a confidence from the share of the prefix observed as VPN exits. Residential proxy exits are never interpolated: where the prefix holds those, the answer — or, without VPN exits, the “not observed” — carries a network context with the services and their exit count. See Methods and evidence.

Who held the address

Where several providers used an address, every source contributes the days it saw each one — a day our probing exited through it, the enriched snapshot’s last sighting, each scan run that found the operator’s certificate — and each day counts half as much for every 30 days before today. Days, not events: probing one provider more often does not win it the address. The leader holds the largest share; where the most recent sighting belongs to someone else, the page says most recently seen as that provider, because a handover takes weeks to show in the shares.

History, week by week

Every address page has a History: each week since probing began, Monday to Sunday, with the strongest evidence that week holds and the providers behind it — for an address one provider held as much as for one that changed hands. First and last seen are per provider there, not only for the address as a whole.

EvidenceThat week
ProbeOur probing exited through the address. Shares count days seen, recent days more
ScanA scan run found a VPN operator’s certificate on it
InterpolatedNeither, but probing reached other addresses of its announced prefix that week. Shares count the prefix’s addresses per provider that week
—Nothing. Consecutive empty weeks are folded into one row

Choose the window on the strip of weeks — drag it, or pick 4, 13 or 52 weeks — and filter by evidence; the choices combine. A week is the smallest unit. An address we never observed still shows its network, prefix and country from our own tables.

Network and service

Two questions, answered apart. What kind of network the address sits in (infrastructure) comes from the network: a datacenter or cloud list that covers it, else the type its operator filed with PeeringDB where that type is unambiguous — Content is a datacenter network, Cable/DSL/ISP an access network for homes and businesses. Everything else is unknown: PeeringDB’s NSP and Enterprise are filed by incumbent carriers and hosting companies alike. infrastructureBasis names the source in one sentence. What kind of service we measured there (serviceCategory) comes from the measurement: a residential or datacenter proxy pool, otherwise a VPN.

A residential proxy exit in a datacenter network is therefore shown as exactly that, not as residential. Until 22 November 2026 the network followed the service, and most residential proxy exits in AWS, Google Cloud or hosting networks read as residential.

Sharing

sharing.iocValue states whether the address identifies the actor or only the service, and sharing.basis the rule that applied. A VPN exit, a datacenter proxy, a Tor exit and an iCloud Private Relay egress are shared by many at once: a weak indicator regardless of attribution confidence (HIGH / LOW). A residential proxy exit on a home or unknown network is a household connection rented out through the pool; how many use it is not measured (UNKNOWN / MEDIUM). A datacenter address with no anonymising service known on it usually serves one tenant (LOW / HIGH). Where no rule applies there is no sharing at all, rather than a guess.