What we keep, and why

Privacy policy

To work for you, the Kringls need to see the conversation, the code and sometimes the server. This page says what is kept, why, with whom and for how long.

Last updated:

On this page
  1. 1Who is responsible
  2. 2What we collect
  3. 3Why we use it
  4. 4AI and your data
  5. 5Your users' data
  6. 6Who we share it with
  7. 7Gmail connection (Google user data)
  8. 8Cookies and measurement
  9. 9How long we keep it
  10. 10Security
  11. 11Your rights
  12. 12Children
  13. 13Changes to this policy

In short

  • We keep what the work needs: your account, conversations, files and the connections you gave.
  • A conversation goes to the model provider that runs the Kringl, mostly Anthropic. We do not train models on your content.
  • We do not sell personal information. It goes only to the providers listed here, to run the service.
  • The site kringl.ai measures visits and enquiries with Wizzo's own tools. The landing page of our Facebook and Instagram campaign also measures with the Meta pixel. The full list is in the section on cookies.
  • To see, correct, copy or delete your data, write to info@wizzo.co.il and we answer within 30 days.

The summary helps you find your way. The full document below is the binding text.

Who is responsible

Wizzo Software Ltd., which runs Kringl, is responsible for the data collected in the service. Address: 3 Totzeret HaAretz St., Petah Tikva, Israel. Any privacy question: info@wizzo.co.il.

This policy covers the site kringl.ai, the app at web.kringl.ai, the Android and iPhone apps and the desktop program. The browser extension has its own privacy page.

What we collect

What is collected, by kind
KindWhat exactlyFrom where
AccountName, email, phone, and a password kept only as a hashFrom you, at sign-up
Workspace and paymentBusiness name, people and number of seats, a card charging token, last 4 digits, expiry and brand, charge log and invoice detailsFrom you and the payment processor
ConversationsMessages, voice notes, files and images you upload, and what the Kringls wrote and capturedFrom you and the Kringls
WorkA copy of the repositories you connected, the Kringls' work logs, and the records of changes made to databasesFrom the connections you gave
Connections and secretsTokens for GitHub, Google and the stores, and passwords and keys you saved in the vaultFrom you
Desktop programContent from the folders you shared, only while a Kringl works on them and only what it needsFrom your computer, with your approval
TechnicalIP address, device and browser type, server logs, a push token for your phone, and crash reports from the appAutomatically
Enquiry from the siteName, phone, email and what you wrote, where you came from (including a campaign ID, if any), and the consent you gaveFrom you, in the form

Your phone number is not verified, and is used to contact you.

Required or optional. You are under no legal duty to give us personal data, and giving it is up to you. A name, an email address, a phone number and a password are needed to open an account and enter into the agreement with us, and without them you cannot sign up. Without a payment method, access stops at the end of the 30 day trial. Connections (a repository, a server, a store, a mailbox) are optional, and without them the Kringls do not work in that place. An enquiry on the site is optional too.

Why we use it

Purposes and their legal basis
PurposeBasis
Providing the service: running the Kringls on your conversation, code and connectionsOur contract with you
Billing, invoices and records under tax lawOur contract with you, and a legal duty
Security, preventing abuse and fixing faultsOur legitimate interest
Improving the service from usage and fault dataOur legitimate interest
Getting back to you after an enquiry from the siteYour consent
Measuring visits to kringl.aiOur legitimate interest, and consent where the law requires it

AI and your data

When a Kringl works, what it needs for the work (the message, pieces of code, files) is sent to the model provider that runs it. That is part of the work, not another use.

We do not train models on your content. What a model provider may do with the data is set by its own terms. When a workspace connects its own model account (for example a Claude subscription), that account's terms and settings also decide whether data is used for training. Check them in the account itself.

The Kringls learn from the work: a lesson found in a project is saved in your workspace's lessons store, so the same mistake is not repeated. The store belongs to the workspace and is not shared with other workspaces, and the workspace admin can switch it off in settings.

Your users' data

When the Kringls work on your product, they may see data of its users, for example in the database or in server logs. You are responsible for that data, and we process it for you, only for the work you asked for.

The Kringls do not send messages, emails or notifications to your users, or delete their data, unless you asked.

Who we share it with

We do not sell personal information. It goes only to the providers in the table, each for what is written next to it.

Providers that process data for us
ProviderWhat forWhere
AnthropicThe models (Claude) that run most of the KringlsUSA
OpenAISpeech to text, reading aloud, transcribing audio and video files, and the Codex engine when chosenUSA
GoogleGemini for the lessons store and the Kringls' memory, Google sign-in, phone notifications (Firebase) and fontsUSA and elsewhere
Cursor, xAI, FireworksAdditional model engines, only when the workspace chooses themUSA
PerplexityWeb search when a Kringl checks a current factUSA
GitHubYour repositories, when you connect themUSA
PayMeCard paymentsIsrael
Morning, iCountInvoicesIsrael
HetznerThe service's servers and the managed serversFinland and Germany
CloudflareDNS and mail forwarding for our domainsGlobal
JetserverHosting of the site kringl.aiIsrael
MetaThe pixel on the campaign landing page (/he/atar) only: measuring the ads on Facebook and InstagramUSA and Ireland
Wizzo's own servicesWizzo Stream (the gateway to model providers), WizzoNow (live updates), WizzoAuth (sign-in codes), POOSH (notifications), Pilot (the chat on the site), Target and WizzoRadar (measuring the site), TODO and POOLSE (tasks and enquiries)Wizzo's servers

Data goes to providers only to run the service. Wizzo is in Israel, and transfers from the European Union to Israel are covered by the European Commission's adequacy decision. Hetzner's servers are in the Union (Finland and Germany). Providers in other countries, mainly the United States, receive data under the safeguards in their data processing terms: the EU standard contractual clauses (SCC) or the EU-US Data Privacy Framework (DPF), depending on the provider. You can ask for a copy of these safeguards at info@wizzo.co.il.

Beyond that, data is disclosed only when the law requires it, or by order of a competent authority.

Gmail connection (Google user data)

Anyone who chooses to can connect their Gmail inbox to Kringl from their personal area. The connection goes through Google's consent screen, is personal to each user, and can be disconnected in the same place at any moment. Kringl does not touch an inbox the user did not connect themselves.

  • What data. With the read scope (gmail.readonly): messages, threads and labels in the user's inbox, according to what they ask in the conversation. With the compose scope (gmail.compose): creating drafts. Kringl does not send email on the user's behalf, does not delete messages and does not change account settings.
  • Why. Only to answer the user about their own inbox and to prepare drafts that they check and send themselves. Not for advertising, not for marketing profiles, not for sale and not for training models.
  • With whom. The message content relevant to the question is sent to the model provider running the conversation, only to produce the answer. By default that is Anthropic (Claude). When the workspace chose another engine, the provider is OpenAI (Codex), xAI (Grok) or Cursor, together with the model providers Cursor works through (for example Fireworks). When the workspace connected its own account or subscription with such a provider, the content goes through that account and under its terms. The access to the inbox itself runs against Google's Gmail server with the user's token. The data goes to no one else.
  • Who reads it. No person at Wizzo reads the content of the email, except when: the user explicitly asked us to look at specific messages; it is needed for security, for example to investigate a bug or abuse; the law requires it; or the data is aggregated and anonymised, so no person can be identified from it, for internal operations.
  • How it is protected. Access tokens are stored encrypted in the database, separately for each user, and are used only for actions the user asked for. All traffic is encrypted with HTTPS.
  • For how long. Inbox content is not stored as a collection: it is read to produce the answer, and only the answer stays in the conversation history. Disconnecting revokes the token at Google and deletes it on our side at once. To delete a conversation or the whole account, write to info@wizzo.co.il.

Limited Use statement, our own and in Google's wording:

Kringl's use and transfer to any other app of information received from Google APIs will adhere to the Google API Services User Data Policy, including the Limited Use requirements. Raw or derived user data received from Workspace APIs is not used to develop, improve or train generalized AI or ML models.
The use of information received from Google Workspace scopes will adhere to the Google User Data Policy, including the Limited Use requirements.

Cookies and measurement

The site and the app keep what is in the table in your browser. The site's measuring tools are Wizzo's own, not Google Analytics. The one exception is the landing page of our Facebook and Instagram campaign (/he/atar): only there the Meta pixel runs, telling Meta that the page was opened and that a form was sent, to measure the ads and show them to people who may be interested. Meta receives the page address, the IP address, browser details and its own cookies, not the name, phone or link you typed in the form. What Meta does with it is in Meta's own privacy policy.

What is kept in your browser
NameWhereWhat forHow long
portal_token and the workspace cookieweb.kringl.aiKeeping you signed in and remembering your workspace1 year
wizzo_a11y_v1kringl.aiRemembering the settings you chose in the accessibility menuUntil deleted
_tuidkringl.ai (Target)Recognising a returning browser, and knowing where visits and enquiries came from1 year
The WizzoRadar IDkringl.aiVisit statistics: pages, referrer, page speed and click positions for a heatmap2 years, plus a visit ID until the browser closes
Pilotkringl.aiThe chat window on the siteUntil deleted
poosh_regkringl.aiSigning up for browser notifications, only if you allowed themUp to 30 days, renewed
_fbp, _fbc (the Meta pixel)kringl.ai/he/atar onlyMeasuring how many people came from the ads and how many left their details90 days

You can block or delete cookies and local storage in your browser settings. The site works without the measuring tools; in the app you cannot stay signed in without them.

How long we keep it

Retention periods
WhatHow long
Account, conversations, files and the Kringls' memoryWhile the account is active, and up to 30 days after a deletion request
The Kringls' full work logs180 days
Server logs14 days
BackupsRotating backups, the oldest at most six months old
A voice recording in a conversationTurned into text; the recording is not kept. A voice note in the team box is kept like a file
Billing records and invoices7 years, as tax law requires
A managed server that was shut downA last backup for 30 days
An enquiry from the siteUntil you ask us to remove it

A deletion request removes the data from the service. Whatever is already in a backup is removed when that backup rotates out.

Security

All traffic is encrypted with HTTPS. Passwords are kept only as a hash. Connection tokens, and passwords and keys in a project's vault, are stored encrypted. Wizzo staff may access data only to run, support and secure the service. The content of a connected Gmail inbox is not read by a person, except as described in the Gmail connection section.

Please note: a file or image uploaded to a conversation is stored at a long address that is hard to guess, with no password. Anyone who gets the link can open it, so do not pass on a link to a sensitive file.

No method guarantees full security. If a security incident affects your data, we will tell you and the authorities as the law requires.

Your rights

You may see the data kept about you, correct it, ask us to delete it and get a copy of it. You may also ask us to stop marketing to you, and withdraw a consent you gave.

If you live in the European Union you may also restrict processing, object to processing based on legitimate interest, and receive your data in a portable format.

Write to info@wizzo.co.il and we answer within 30 days. Deleting an account or a conversation is done by such a request for now, not by a button in the app.

You may also complain to the Privacy Protection Authority in Israel, or to the data protection authority of your country in the European Union.

Children

Kringl is a work tool for businesses and is not meant for anyone under 18. We do not knowingly collect data about children, and such data that reaches us is deleted.

Changes to this policy

When this policy changes, the date at the top changes. We announce a material change by email or in the app before it applies.

A question about this document? Write to info@wizzo.co.il and a person answers.