Privacy policy
This policy says what the opdns service and the opdns self-hosted node do with data, down to fields and retention periods. You should be able to check every statement here against the code in the public repository. Where the code and this policy disagree, that is a bug, and we fix whichever one is wrong.
1. Who we are
Section titled “1. Who we are”opdns is operated by Solutions Opero Inc, a company incorporated in
Québec, Canada (“we”). Contact for anything in this
policy: [email protected]. We are the controller
for account data and for query logs stored in our cloud. For a self-hosted
node, you are the controller of what your node stores; we are not.
The person in charge of the protection of personal information at
Solutions Opero Inc is, as Québec law provides, the person with the
highest authority in the company. Write to them at [email protected] with
any question, request or complaint about your personal information
(section 9). This policy is also our governance policy for personal
information: who is responsible (this section), why and how we collect it
and who has access (section 18), how long we keep it and how it is
destroyed (section 6), confidentiality incidents (section 14), and how to
make a request or a complaint (section 9).
2. How a query flows
Section titled “2. How a query flows”A DNS query sent to an opdns address reaches one of our points of presence
(PoPs) by anycast. On the PoP, the opdns-edge process identifies the
profile, applies its filtering policy, answers blocked names itself, and
forwards allowed names to an Unbound recursive resolver on the same machine.
Unbound never sees blocked queries and never sees who asked: it receives
queries only from the edge, over loopback.
The edge identifies a profile from, in order: the DoH URL path
(https://dns.opdns.net/<profile>/<device>), the TLS server name for DoT and
DoQ (<device>-<profile>.dns.opdns.net), the destination IPv6 address for
plain DNS over IPv6, or the source IPv4 address if it is a linked IP on a
profile. A query that matches none of these is unidentified.
3. Unidentified queries (the public resolver path)
Section titled “3. Unidentified queries (the public resolver path)”We answer plain DNS sent to our IPv4 addresses from an address that is not linked to a profile without filtering, like any public resolver.
- We create, store and transmit no per-query record, with one narrow exception: a short packet capture during a flood against one of our servers (below).
- We keep aggregate counters only: queries per PoP by transport,
identification method (
nonefor this path) and outcome, as Prometheus metrics. The counters contain no IP addresses and no domain names. - To enforce rate limits, the edge keeps a token bucket per source address (IPv4 /32 or IPv6 /64) in memory only. It forgets idle buckets, caps the table, and never writes it to disk or sends it anywhere.
- Operator-level blocks (see our operator blocks and takedowns policy) apply on this path too.
A query that names a malformed or unknown profile id (or one that was deleted) is treated as carrying no profile: it takes this public path, unfiltered, and we do not log it.
Packet captures during a flood. When one of our servers detects a flood against its own address (the server’s own unicast address, not our shared anycast addresses), it may keep a short packet capture, up to 60 seconds and at most one an hour, of the DNS traffic to that address: the queries, including their source IP addresses and query names, and our answers. The capture is stored only on that server, readable only by our staff, never copied off the server automatically, and deleted automatically after 7 days. We use it only to identify and report abusive sources.
4. Identified queries: what is logged
Section titled “4. Identified queries: what is logged”For an identified query whose profile has logs on, the edge produces one record with these fields:
| Field | Content |
|---|---|
ts |
time the query was received (microseconds, UTC) |
pop, node |
which PoP and server answered |
profile_id |
the 6-character profile id |
device_id, device_name |
the device name you put in the DoH path or DoT/DoQ name (sanitised, length-capped), and a stable id derived from it; empty for plain DNS |
client_ip |
the source address of the query; zeroed on the PoP when client-IP logging is off |
transport |
Do53 UDP/TCP, DoT, DoH, DoQ or DoH3 |
qname, qtype |
the domain name and record type asked for |
status |
allowed, blocked, rewritten or error |
reason, list_id |
the list or your own rule that blocked or rewrote the query |
rcode |
the DNS response code |
dnssec_validated |
whether the answer was DNSSEC-validated |
cache_hit |
whether the answer probably came from cache (a latency heuristic) |
latency_us |
time to answer |
answer_ips |
at most the first two A/AAAA addresses in the answer, for the destinations view |
We do not log the full answer, EDNS options, TLS client data, the HTTP headers or cookies of DoH requests, or anything about the contents of your traffic beyond the DNS question and the first answer addresses.
5. What your profile settings change, and where
Section titled “5. What your profile settings change, and where”Each profile has the settings below. Three of them take effect on the PoP, before anything leaves the machine. The others take effect at ingest in our control plane. The difference matters: data covered by an ingest-side setting crosses our internal message bus before the setting applies.
| Setting | Effect | Applied |
|---|---|---|
| Logs off | The edge produces no per-query record, only hourly per-profile totals (below). | PoP |
| Client IP off | client_ip is removed before the record leaves the PoP. |
PoP |
| Domain logging off | qname, answer_ips and rule-text reasons are blanked before the record leaves the PoP. Status, device, list name and timing are kept. |
PoP |
| Retention (1 to 90 days) | Each stored row gets an expiry time, and ClickHouse deletes the row after it. Unset means 30 days; a value above 90 is cut to 90. | ingest |
Destination cloud |
Records are stored in our ClickHouse. | ingest |
Destination self-hosted |
Records are queued for your self-hosted node and are not stored in ClickHouse. | ingest |
Destination both |
Both of the above. | ingest |
Destination none |
Records are discarded at ingest. | ingest |
| Storage region | Where records stored in the cloud are kept. Canada is the only region today and the default; EU storage is announced and cannot be chosen yet. | ingest |
Defaults for a new profile
Section titled “Defaults for a new profile”A profile you create starts with these settings. You can change each of them at any time; a change applies to queries from then on.
| Setting | Default |
|---|---|
| Logs | on |
| Client IP logging | on |
| Domain logging | on |
| Retention | 30 days (you can choose 1 to 90) |
| Destination | cloud |
| Storage region | Canada |
| Queue for a self-hosted node | 24 hours (applies only with destination self-hosted or both) |
| EDNS Client Subnet | anonymised: authoritative servers see your network truncated to /24 (IPv4) or /56 (IPv6), never your full address (section 16) |
| Blocked answers | NXDOMAIN |
| Blocklists | step 3 (Balanced) of the Privacy protection slider; every device telemetry switch off |
| Security protections | threat intelligence, cryptojacking, DNS rebinding and the CSAM switch on; newly registered domains off. IDN homograph, typosquatting and algorithm-generated-domain protection are off and hidden until the resolver implements them |
| Parental controls | none |
These are the values the control plane writes when it creates a profile, quoted verbatim as the API returns them for a new profile:
{ "security": { "rebinding": true, "idn_homograph": false, "typosquat": false, "nrd": false, "dga": false, "cryptojacking": true, "threat_intel": true, "csam": true }, "settings": { "block_mode": "nxdomain", "logs_enabled": true, "log_client_ip": true, "log_domains": true, "log_destination": "cloud", "retention_days": 30, "cname_uncloak": true, "cache_boost_ttl": 0, "ecs": "anonymised", "node_queue_hours": 24, "log_region": "ca" }}An automated test fails our build when this block and the code disagree.
Per-profile counters: we keep hourly per-profile totals (total, allowed, blocked, rewritten, refused, errors) for 400 days, so the overview page works with logs off. For a profile with logs on they are counted from the stored records. For a profile with logs off, the PoP sends one count per profile and hour instead of records; the counts carry no domain names, IP addresses or device names.
6. Where data lives and for how long
Section titled “6. Where data lives and for how long”| Store | What | Where | Retention |
|---|---|---|---|
| Edge memory | the query while it is answered; rate-limit buckets | every PoP | seconds; buckets until idle |
| Edge disk spool | records waiting while the message bus is unreachable | the PoP that answered | until delivered; capped at 1 GiB, oldest dropped first |
logs stream (NATS JetStream) |
records after the PoP-side settings, before the ingest-side settings | Canada | 24 hours maximum |
ClickHouse queries |
records after all settings, for destination cloud or both | Canada | your retention setting |
| ClickHouse hourly rollups (status, domains, devices, destinations) | counts derived from stored rows | same | expire with the rows they summarise |
ClickHouse aggregates |
hourly per-profile totals, with no names or IPs | same | 400 days |
nodelogs stream |
records queued for your self-hosted node | Canada | until your node acknowledges them; at most your node queue setting (24 hours by default, or 12 hours, 1 hour, or not queued), never more than 24 hours, and 1 GiB fleet-wide |
| Postgres | account, profiles, rules, devices, linked IPs, node records, sessions (with user agent and IP), API token hashes, audit log | Canada | while the account exists; see section 9 |
| Postgres mail events | bounce and complaint notices from our mail provider: your email address, the kind of event and the provider’s reason | Canada | 90 days, or until the account is deleted, whichever comes first |
| Postgres suppression list | email addresses that hard-bounced, complained or kept soft-bouncing, with the reason | Canada | kept after the account is deleted, so we never mail that address again; ask [email protected] to remove yours |
| Operational logs | service logs and API HTTP request metadata | Canada | 14 days; DNS query logs are never written here |
| Flood packet captures (section 3) | up to 60 seconds of DNS traffic to one server’s own address during a detected flood: source IP addresses, query names, our answers | the server that was flooded | 7 days |
| Backups | Postgres (continuous, point-in-time, plus hourly encrypted copies) and ClickHouse (daily encrypted copies) | Canada | Postgres 30 days; ClickHouse: every copy for 14 days, then the first copy of each month for 120 days |
Storage region. Query logs we store in the cloud (ClickHouse queries,
its rollups and aggregates) are kept in Canada. Each profile has a storage region setting: Canada is
the only region today, and the default. EU storage is coming: once it
exists you will be able to move a profile’s logs there, and we will update
this section before it is offered.
PoPs are servers we rent in the United States and Germany; the status page lists each one. A query and its record exist on the PoP that answered, in that country, until they are sent to the control plane in Canada, which normally happens within a second.
When a period above ends, we destroy the data: rows are deleted or expire in the databases, the streams and the spool, and backup copies expire on their own schedule. What we keep after you delete your account is listed in section 9: each profile’s id, so that it is never reused; audit entries with your identity replaced by a one-way hash; and, if your address is on it, the suppression list entry. Aggregate statistics that contain no profile ids, names or addresses (section 8) identify no one and are not covered by these periods. Some of these places are outside Québec (section 11).
7. Self-hosted nodes
Section titled “7. Self-hosted nodes”The self-hosted node is the same resolver, running on your hardware.
- Queries your node resolves never reach us. Logs your node produces stay in its SQLite file. They leave the node only as answers to your own dashboard queries (described below).
- Enrolled mode. The node keeps one outbound WebSocket open to
link.opdns.net. Over it we learn the node’s id, versions, uptime, last sync time, clock offset, whether its Unbound is healthy, the link’s round trip, the ids of the profile-signing keys it trusts, its log health (SQLite file size, oldest record time, pending and dropped record counts) and connection times, and, as with any connection, the node’s public IP address. We do not store that address: we store the approximate location looked up from it (country, and city when known), shown only to you on the node’s entry. We send it notices of profile and list updates. For profiles whose destination isself-hostedorboth, we also send it the logs our cloud PoPs generated for that profile. - Logs for a self-hosted destination still pass through us. When your
devices send queries to our cloud addresses, our PoPs answer them and log
them as described in section 4. The records travel on the
logsstream, and we hold them in a per-profile queue until your node acknowledges them (at most 24 hours; you can choose 12 hours, 1 hour, or not to queue them at all, in which case records made while the node is offline are dropped). If the node stays offline longer, the undelivered records are lost. They are not moved to ClickHouse. - Dashboard queries relay through the cloud. When you open logs or analytics in the cloud dashboard for a profile whose logs live on a node, the request goes to the node over the link, and the result comes back through our servers to your browser. TLS protects each hop, but our servers can read the result in transit. We do not store it.
- Standalone mode. No account and no link. We see only downloads of the public list artifact (IP address and time), which stay in operational logs for 14 days.
The self-hosted documentation arranges these facts by mode, with how to remove each: Privacy of your node, and Self-hosted overview for installing and enrolling a node.
8. Account data
Section titled “8. Account data”We hold:
- your email address and the id of your op account;
- sessions: hashed token, user agent, IP address, expiry (30 days) and a reference to the op account sign-in that started it;
- API tokens: hash, name, scopes and last use;
- profile configuration, including linked IPs and DDNS tokens;
- for a paid plan: the plan, the ids of your customer record and subscription at Stripe, the subscription’s status and renewal date, and the record of each billing notice Stripe sent us;
- an audit log of sign-ins and sensitive changes.
You sign in with your op account (opauth), which holds your password, passkeys, second factors and recovery codes; opdns never sees them. It tells us your email address, whether it is verified, and when you signed in.
We send transactional email (waitlist, notices, alerts) through Resend. We send no marketing email without opt-in.
Paid plans are sold by Solutions Opero Inc: you buy the plan from us.
Stripe (Stripe Payments Canada, Ltd., with Stripe, LLC in the United
States) processes the payment for us: its checkout takes the payment, it
computes the sales tax we add (GST/HST and QST in Canada), and it issues
the invoices and receipts in our name. Stripe holds your card, your billing
name and address, and the tax id if you give one; we keep only what is
listed above. We give Stripe your email address and your organisation’s id
when you start a checkout. Your card statement shows OPDNS.
We do not sell or share personal data, use it for advertising, or give it to data brokers. We use query logs only to show them to you, to operate and secure the service, and for aggregate statistics that contain no profile ids, names or addresses.
9. Your rights and how to exercise them
Section titled “9. Your rights and how to exercise them”Depending on where you live (GDPR, UK GDPR, CCPA/CPRA, PIPEDA, Québec’s Act respecting the protection of personal information in the private sector, and others), you have the right to access, correct, export and delete your data, to withdraw your consent, and to object to processing:
- Access and export: in the dashboard (Account → Export your data)
or through the API (
POST /v1/auth/me/export). We build an archive of your account, sessions, API token metadata, audit entries, and for each profile its settings, rules, linked IPs, devices, nodes and the query logs we store in the cloud within their retention, and email you a link to it. The link and the archive expire after 24 hours. The archive is a zip file of JSON, NDJSON and CSV files: structured, commonly used formats that other software can read. On request, we send it instead to a person or body that the law authorises to collect it. - Correction: edit the data in the dashboard (your email address in
your op account). For anything you cannot
edit there, write to
[email protected]and we correct what is inaccurate, incomplete or equivocal. - Withdrawing consent: turn logs, client-IP logging or domain logging off (section 5), remove a linked IP, revoke a node, or delete the profile or the account. A change applies to queries from then on. Some data is needed to provide the service at all (an account needs an email address); withdrawing it means closing the account.
- Deletion of logs: lower the retention or delete the profile. See our data rights process for exactly what each action removes.
- Deletion of the account: in your op account (the dashboard’s Account → Delete account links there), which tells us. Sessions, API tokens and nodes are revoked and your profiles stop filtering at once (devices that still use them get unfiltered, unlogged answers). You can cancel there until the deletion is due (7 days when we start it on your emailed request); then we delete the account’s rows, scrub its profiles to their ids (so an id is never reused), replace your identity in the audit log with a one-way hash, purge the queues for your self-hosted nodes and delete every query log row and count of your profiles in ClickHouse. The bounce and complaint events of your address are deleted too; if your address is on the suppression list (section 6) it stays there.
- Ceasing dissemination and de-indexing: we publish no personal information about you. If you think that we disseminate information about you, or that a hyperlink attached to your name gives access to it, contrary to the law or a court order, or in a way that seriously injures your reputation or privacy, you may ask us to stop disseminating it or to de-index the link.
- Decisions made by automated processing: section 20 lists them, and how to ask about one and have a person review it.
- Requests by email: if you cannot log in, or for anything the
dashboard does not do, write to
[email protected]. We verify your identity before acting and answer in writing within 30 days of receiving the request. If we refuse all or part of it, we say why and on which provision of the law, and how to contest the refusal.
Nothing is deleted inside a backup: deleted data leaves our backups when those backups expire (section 6), at most 30 days later for account data and 120 days later for query logs. If we ever restore a backup within that window, we repeat the deletion before the restored system serves anyone.
Complaints. Write to the person in charge of the protection of personal information (section 1); we answer within 30 days. You may also complain to your data protection authority at any time. In Québec, that is the Commission d’accès à l’information du Québec (https://www.cai.gouv.qc.ca). If we refuse a request for access or correction, or do not answer within 30 days, you may ask the Commission to examine the disagreement within 30 days of the refusal or of the end of that period.
10. Lawful basis (GDPR)
Section titled “10. Lawful basis (GDPR)”- Contract: resolving your queries, storing the logs you asked us to keep, and account data.
- Legitimate interest: abuse prevention (rate-limit state, operator blocks, operational logs, flood packet captures) and security.
- Legal obligation: responding to valid legal process (counted in the transparency report).
11. Sub-processors
Section titled “11. Sub-processors”| Vendor | Purpose | Data | Region |
|---|---|---|---|
| hosting provider (Canada) | control plane, databases, object storage, backups | all cloud data | Canada |
| network provider for points of presence | PoP servers and IP transit | queries and records in transit on each PoP | United States, Germany |
| Resend | transactional email | email address and message content | not yet confirmed |
| Stripe Payments Canada, Ltd., with Stripe, LLC, 354 Oyster Point Boulevard, South San Francisco, CA 94080, United States (Stripe) | payment processing for the paid plans: checkout, card payment, sales tax calculation, invoices and receipts issued in our name, the billing portal | email address, organisation id, billing name and address, the card (held by Stripe), the tax id if you give one, subscriptions, invoices and credit notes | Canada and the United States |
The status page at opdns.fail collects no personal data: it offers a feed
and no email subscription.
Communication outside Québec. Your personal information may be communicated outside Québec: to our hosting in Canada, to the PoPs in the United States and Germany (section 6), and to the vendors above, in the countries listed. Before we communicate personal information outside Québec, we assess the privacy impact: how sensitive the information is, what it is used for, the protection measures (contractual ones included) that will apply to it, and the legal framework of the place where it goes. We communicate it only if that assessment shows it will be adequately protected, under a written agreement with the recipient.
We announce changes on this page 30 days in advance.
12. Cookies
Section titled “12. Cookies”The dashboard sets one first-party session cookie, which you need to stay logged in. We use no analytics, advertising or third-party cookies, and no third-party scripts or fonts on our sites.
13. Children
Section titled “13. Children”opdns is not directed at children, and accounts are for adults (18 or older, terms section 2). We do not knowingly collect personal information from a child. Adults configure parental controls for devices that children use. Queries from those devices are logged according to the profile’s settings, which the account holder controls; logs are on for a new profile (section 5).
In Québec, personal information about a minor under 14 may be collected
only with the consent of the person having parental authority or the
tutor. If a device on one of your profiles is used by a child under 14,
you must be that person, or have their consent, before the profile logs
the child’s queries; otherwise turn logs off for that profile. If you
think we hold information about a child without that consent, write to
[email protected] and we delete it.
14. Security
Section titled “14. Security”- TLS on every external connection.
- Tokens are stored as hashes; opdns holds no passwords (your op account does, hashed).
- Only the operator has access to production (section 18).
- List artifacts and profile data are signed or authenticated.
Report problems as described in our security policy.
Confidentiality incidents. A confidentiality incident is any access to, use, communication or loss of personal information that the law does not authorise. We record every one in a register of incidents. When an incident presents a risk of serious injury to the people concerned, we notify the Commission d’accès à l’information du Québec and those people promptly, and the authorities and people that the law of other places requires (for example the GDPR or PIPEDA). We take reasonable measures to reduce the risk of injury and to prevent a new incident of the same kind.
15. Changes
Section titled “15. Changes”A dated changelog sits at the bottom of this page. We email account holders about material changes 30 days before they take effect.
16. EDNS Client Subnet
Section titled “16. EDNS Client Subnet”EDNS Client Subnet (ECS) is anonymised by default for new profiles; you can turn it off. “Anonymised” sends authoritative DNS servers your network truncated to /24 (IPv4) or /56 (IPv6). “Full” sends your IP address itself (by default the complete address; the operator may limit IPv6 to /64) to every authoritative server queried on your behalf, which those operators may log; choose it only if more precise CDN steering is worth that disclosure. Private-network addresses are never sent, a client’s own ECS opt-out is always honoured, and queries without a profile never carry ECS. With “Anonymised”, answers tailored to a network are cached per truncated prefix, so every client of the same /24 or /56 shares them. Self-hosted nodes never send ECS.
17. Web3 names
Section titled “17. Web3 names”Resolving Web3 names (ENS .eth, and Unstoppable Domains and Handshake names where the operator configures them) is off by default and turned on per profile. When it is on, a Web3 name is answered through a third-party gateway: an ENS name name.eth becomes an alias of name.eth.limo (the eth.limo gateway by default), so the gateway’s DNS servers and, when you open the site, its web servers see the name, as they would for any website; a Handshake name is asked of the operator’s Handshake resolver, which sees the name and the PoP’s address but never your address or network. opdns runs no blockchain client and sends nothing else about you to these services.
18. Why we collect personal information, how, and who sees it
Section titled “18. Why we collect personal information, how, and who sees it”We collect personal information only for these purposes:
- to resolve and filter DNS queries as your profiles say (sections 2 to 5);
- to store and show you the logs and statistics you choose to keep (sections 4 to 7);
- to run your account: sign-in, security, notices, plans and billing (section 8);
- to secure the service and prevent abuse: rate limits, operator and source blocks, flood packet captures, operational logs and the audit log (sections 3, 6 and 20);
- to answer your requests, and legal process (section 10).
We collect it by these means:
- from you, when you sign up, configure profiles, and use the dashboard or the API;
- from your devices, when they send DNS queries to our addresses: the query, its source address and the device name you put in the address (sections 2 and 4);
- from your self-hosted node, when it is enrolled, and from a geolocation database, for its approximate location (section 7);
- from Stripe, about your subscription (section 8), and from our mail provider, about bounces and complaints (section 6).
We do not buy personal information or collect it from any other source.
Inside Solutions Opero Inc, only the people who operate the service have access to personal information: to the production systems, databases and backups, and to the support, privacy, abuse and billing requests, each only as far as their task needs. Outside the company, the sub-processors of section 11 handle it only for the service each provides, and we disclose it to authorities only under valid legal process (section 10).
19. Technologies that identify, locate or profile, and privacy by default
Section titled “19. Technologies that identify, locate or profile, and privacy by default”Query logs can show what a device does on the internet, and the source address of a query (the client IP) can indicate roughly where the device is. Query logging, client-IP logging and domain logging are therefore settings of each profile that you control (section 5):
- For a new profile they are on: each query is logged with its client IP and domain name and kept 30 days in our cloud in Canada (section 5, “Defaults for a new profile”). These are not the most private settings.
- The most private choice is logs off: we then keep only hourly totals, with no domain names, IP addresses or device names. With logs on, you can turn client-IP logging and domain logging off, shorten the retention to 1 day, send logs only to your self-hosted node, and turn EDNS Client Subnet off (section 16).
- You change each in the profile’s settings in the dashboard or through the API, at any time; a change applies to queries from then on.
Linked IPs and device names identify your devices so that your profile applies to them; you add them yourself. For an enrolled self-hosted node, we look up the approximate location (country, and city when known) of its public IP address and show it only to you (section 7); this is done for every enrolled node.
We do not profile you: we draw no conclusions about you from your queries, use no analytics or advertising technology on our sites or in the dashboard (section 12), and sell nothing about you (section 8). The analytics in the dashboard are your own logs, summarised for you.
20. Decisions made by automated processing
Section titled “20. Decisions made by automated processing”These are the only decisions we make by automated processing alone, without a person looking. Each follows a fixed rule:
- Query rate limits. Every source address has a query budget per second (terms section 4); queries over it are dropped or answered truncated. The state is kept in memory only (section 3).
- API limits. Requests made with an API token past the plan’s limit per minute are refused with HTTP 429, and creating profiles, nodes, tokens or linked addresses past the plan’s limits is refused.
- Over quota. When an organisation passes its plan’s monthly queries, its profiles keep resolving and filtering but logging pauses until the month resets; the owners and admins get an email at 80 % and at 100 %.
- Failed payments. When a renewal of a paid plan is not paid, a 7-day grace starts with two emails; then the organisation moves to the free plan and keeps everything.
Suspending a profile or an account, and blocking a source address on every
PoP, are decided by a person. You can ask [email protected] about any
decision above: we tell you the personal information used, the reasons and
main factors behind it, and your right to have that information corrected,
and a person reviews the decision with your observations.
Changelog
Section titled “Changelog”- 2026-10-21: Initial version.
- 2026-11-01: Section 3 allows a short packet capture of the DNS traffic to one of our servers’ own addresses during a flood against it (up to 60 seconds, kept only on that server, deleted after 7 days, used only to identify and report abusive sources); sections 6, 10 and 18 list it.