Privacy Policy

Last Updated: September 9, 2026

Key changes in this update

  • Email you send to our support and contact addresses may be read and sorted with the help of AI tools we operate, so that a person can reply faster. A person reads every message. We added Anthropic and Google (Gmail) to the sub-processor table and removed an outdated sentence about avoiding AI processing by email. See Sections 2.8, 3, 4, 5, 8 and 17.
  • We now disclose how we read Reddit posts you save. Our own servers read the public post once at save time, identifying as MarkItAgent and never logged in. Because Reddit only serves this to a residential network, the request is routed through a third-party proxy provider (Proxy-Seller, operated by SSV IT Provider Online Services Ltd, Cyprus) that relays it from a residential IP. Only the public post address is sent - never anything about you, and no cookies. Apify remains the fallback. See Sections 2.4 and 4.
  • We now disclose how we read X (Twitter) posts you save. Our own servers read them from X's public syndication endpoint, falling back to the open-source fxtwitter and vxtwitter embed services. Only the public tweet address is sent, never anything about you. See Sections 2.4 and 4.
  • Our own servers now fetch public Instagram, Facebook, TikTok and LinkedIn posts you save. Until now a third-party service (Apify) fetched them for us. From this update, MarkIt fetches the public post directly, once, at the moment you save it, identifying itself as MarkItAgent and never logged in. Apify remains a fallback when our own fetch fails, and still handles YouTube captions. Nothing about you is sent to those platforms: only the public address you saved. See Sections 2.4 and 4.
  • Our own servers now also fetch public Threads and Pinterest posts you save. Threads posts were previously fetched by Apify, which remains the fallback; Pinterest posts are now fetched directly by our servers. As with the platforms above, MarkIt fetches the public post once, at the moment you save it, identifying itself as MarkItAgent and never logged in, and nothing about you is sent - only the public address you saved. See Sections 2.4 and 4.
  • Previous update (September 2, 2026): We now load two third-party analytics tools, and this policy previously said we loaded none.Microsoft Clarity (session recordings) and Google Analytics were added on September 1, 2026 and run on every page, including the signed-in dashboard. Three statements here were made false by that change and are corrected in this update: that we used no third-party analytics, that only strictly necessary cookies were set, and the reason given for showing no consent banner. See Sections 2.7, 4 and 16.
  • You can now report a search that returned the wrong things, and doing so stores the text of that one search in our database and emails it to our internal inbox. It is opt-in per search: we do not record unreported searches into our own database. See Section 2.8, which also notes that the third-party analytics tools added on September 1 receive page addresses, and the search page carries your terms in its address.
  • Previous update (August 13, 2026): Page views are now linked to your account while you are signed in. Previously our analytics used only a random per-tab session ID. They still do while you are signed out, but a page view recorded while you are signed in now stores your user ID as well, so we can see where new people get stuck between signing up and saving their first item. See the rewritten Section 2.7.
  • Added product events (Section 2.7): a short record of things like completing signup, saving an item, or connecting WhatsApp or Telegram. We store the outcome and a reason, never the content - not the item, not the link, not what you searched for.
  • Added campaign parameters (utm_source, utm_medium, utm_campaign) from the address bar when you arrive from a link we shared, so we can tell which posts brought people in.
  • Added page-speed measurements (Section 2.7): anonymous loading-performance numbers from real visits.
  • Added stated deletion schedules for all of this: page views and page-speed measurements after 365 days, product events after 180 days. These now delete automatically rather than being kept indefinitely.
  • Previous update (August 2, 2026): expanded Section 2.7 to cover mouse-movement paths and to note that heatmap analytics run across our public pages as well as the dashboard.
  • Previous update (July 13, 2026): updated Section 2.11 (API access and connected AI clients) to match what the connector can now do: a connected AI client can also save new links and notes into your library at your direction (each genuine new save counts toward your monthly capture limit), cancel a reminder you ask it to cancel (deleting that one reminder; Google-Calendar-synced reminders excluded), and - only when you ask it to - send us a short feedback note, which is handled exactly like widget feedback (Section 2.8): delivered to our internal inbox and auto-classified by OpenAI for triage. It still cannot delete your saved items or change account settings.
  • Extended the feedback disclosures in Sections 2.8 and 5 to cover feedback notes sent through a connected AI client, not just the in-app widget, and corrected the model name used for feedback triage (gpt-5.4-nano).
  • Previous update (July 7, 2026): added Section 2.11 itself (API access and connected AI clients), Cloudflare R2 as the encrypted off-site backup store (Section 4), the widget-feedback OpenAI triage disclosure, the full named sub-processor table, the legal-basis table (Section 3), jurisdiction-specific sections (CCPA/CPRA, UK, EU, Israel), the heatmap-analytics disclosure, the `items.content_en` derived-data disclosure, the DPF-plus-SCCs transfer explanation, and the expanded early-access request disclosure (Section 2.10).

At a glance

MarkIt is a universal capture tool run by Tomer Ezri, a sole proprietor based in Israel. We save the links, notes, screenshots, documents, and messages you send us, organize them with AI, and let you search them later. We do not sell your data, we do not train third-party AI models on it, and we do not show you ads. This policy is the long version. If anything below is unclear, email tomer@mark-it.co.

1. Who we are

  • Controller: Tomer Ezri, sole proprietor, Israel, trading as "MarkIt" at mark-it.co.
  • Primary contact for privacy questions and rights requests: tomer@mark-it.co.
  • Product support: support@mark-it.co.
  • EU representative (Article 27 GDPR): we have not yet appointed one. We will name an EU representative in this policy before the Service is generally available to EU residents outside the closed invite-only early access. In the meantime, EU users can reach us at tomer@mark-it.co.
  • UK representative (Article 27 UK GDPR): same status as the EU representative above.
  • Data Protection Officer: we have not appointed one. We fall below the GDPR and UK GDPR thresholds that would make appointment mandatory. Privacy questions go to tomer@mark-it.co.

2. What data we process

2.1 Account data

  • Email address, display name, and (for password sign-ups) a password hash managed by Supabase Auth.
  • If you sign in with Google, we receive your email, name, and avatar URL from Google.
  • Invite code (during early access).

2.2 Captured content

  • URLs you paste, save, or send to the bots.
  • Text notes you type.
  • Images, screenshots, and handwritten notes you upload.
  • Documents you upload (PDF, DOCX, TXT, MD). We extract the text server-side and store up to 50,000 characters of the extracted text alongside the item.
  • Media you send through the WhatsApp or Telegram bots.

2.3 AI-derived metadata

  • Categories, tags, content-type labels, AI-generated headlines, and summaries our system produces from your items.
  • Text embeddings used to power semantic search.
  • Scraped article text we pull from public web pages you save.

2.4 Integration data

  • Google Calendar: if you connect it, we receive an OAuth access token and refresh token from Google. We store both only in the NextAuth session cookie on your browser. We never write them to our database.
  • WhatsApp: for WhatsApp we store (a) a SHA-256 hash of your phone number used as the primary key in our database, (b) a masked display form (for example, "+972-50-***-1427") for UI purposes, and (c) during the active WhatsApp beta, the raw phone number in one table (user_engagement.phone_raw) used by our outbound-messaging code. We are migrating to hash-only storage; until that migration completes, the raw phone is retained while you have WhatsApp linked and is deleted when you unlink WhatsApp or delete your account. Every inbound and outbound message is logged in an audit table for up to 30 days for abuse-prevention and support purposes. If you unlink WhatsApp we delete the logs for your account immediately.
  • Telegram: we store your numeric Telegram user ID, a short-lived link token used to tie your Telegram account to your MarkIt account, and an audit log of bot messages.

2.5 Engagement data

  • Opt-in status for email (default on) and WhatsApp proactive messages (default off).
  • Timestamps for the last email and WhatsApp messages we sent you, so we can respect daily and monthly frequency caps.
  • An engagement segment (for example, "new", "active", "at risk", "dormant") computed daily from your activity. We use it only to decide cadence, never to gate access or set prices.
  • A per-message record (send time, open, click, reply, unsubscribe, Resend identifier) for every engagement email or WhatsApp we send you.

2.6 Device and session data

  • Supabase authentication cookies (session cookies named with an sb- prefix) so you stay signed in.
  • A NextAuth session cookie, if you connect Google Calendar.
  • A short-lived markit_invite_code cookie (10 minutes, during Google OAuth redirect).
  • A random session identifier stored in your browser's sessionStorage, shared by our in-house heatmap analytics, page views, and page-speed measurements. It is per tab and is cleared when you close it.

2.7 In-house product analytics

Most of our analytics is our own: the heatmaps, page views, product events and page-speed measurements described below are built in-house and stay on our systems. Since September 1, 2026 we also load two third-party analytics tools on every page, described in the next subsection. We do not use PostHog, Mixpanel or Amplitude. Nothing here is shared with an advertising network. Your IP address is not written to our own database, though the third-party tools below receive it as part of an ordinary web request.

Third-party analytics: Microsoft Clarity and Google Analytics

Microsoft Clarity records sessions: it replays how a page was used, including pointer movement, clicks, scrolling and rage-clicks. Google Analytics 4 records page views and interaction events. Both are loaded on every page of MarkIt, including the signed-in dashboard, which means a Clarity session recording can capture the parts of your library visible on screen while you use it. Both set their own cookies in your browser and send data to Microsoft and Google respectively, who process it under their own terms.

These were added on September 1, 2026. An earlier version of this policy said we used no third-party analytics and undertook to show a consent banner before adding any. We are correcting that statement here rather than leaving it standing, and the consent question is described in Section 16.

What we changed on September 2, 2026. Google Analytics now receives page addresses with the query part removed, so the words you type into search are no longer included in what we send it; campaign tags such as utm_source are kept, because that is how we tell which link brought someone here. We also excluded the search results area from Clarity's recordings, so your search terms, the titles of what you saved, and your recent searches are not captured in a replay. One gap remains and we would rather name it than leave it implied: Clarity records the address of each page you visit, and the search page carries your search terms in its address, so those still reach Microsoft. We are deciding how to close that.

Heatmaps: clicks, scrolling, and mouse movement

For each heatmap event we store: a random per-tab session ID (not your user ID, not your IP), the page URL (capped at 500 characters), the event type (click, scroll, mouse movement, rage-click, dead-click, page view), normalized coordinates, mouse-movement paths, scroll depth, time on page, and the element you interacted with (tag name, ID, CSS classes, and up to 50 characters of visible text such as a button label).

We never record the text you type: inputs, textareas, selects, and any field marked as a password, email, or phone number are masked before being written anywhere.

Page views

For each page view we store the page path, the referring website, a country derived from your connection, a random per-tab session ID, and any campaign parameters in the address bar (utm_source,utm_medium, utm_campaign) when you arrive from a link we shared.

If you are signed in, we also store your user ID against that page view. This changed on August 12, 2026; before that, page views were linked only to the random session ID. We made the change because most people who create an account never save a single item, and without knowing which account a visit belongs to we cannot see where they get stuck. Page views recorded while you are signed out are not linked to your account, and we do not retroactively attach them to you if you sign in later.

Product events

We record a short entry when something meaningful happens in the product: completing signup, being turned away because you had no invite code, saving an item, a save failing, and connecting WhatsApp or Telegram. Each entry holds the event name, your user ID where you have an account, the random session ID, the time, and a small set of plain descriptors such as the outcome, the reason a save failed, the sign-in provider, or which messaging channel you connected.

These entries never contain your content. Not the item you saved, not the link, not its title, not the text of a note, and not what you searched for. When someone is turned away at the invite gate we record only that it happened, with the sign-in provider - no email address and no account ID.

Page-speed measurements

We collect standard loading-performance measurements (Core Web Vitals) from real visits: the metric name, its value, a good/poor rating, the page path, the random session ID, and a country. These are not linked to your account.

How long we keep it

Page views and page-speed measurements are deleted after 365 days. Product events are deleted after 180 days. Heatmap data is covered by the schedule in Section 4. These deletions run automatically on a schedule, not on request. If you delete your account, the user ID on any of these records is cleared, so what remains is an unattributed count.

2.8 Communication data

  • Feedback and support emails you send us (including the star rating, message, and the page URL from the widget). Feedback is forwarded to an internal inbox at markit.ai.contact@gmail.com via Resend so we can reply, and the message text is sent to OpenAI (gpt-5.4-nano) to auto-classify its category and priority for triage (see Section 5). A feedback note you ask a connected AI client to send (Section 2.11) is handled the same way: same inbox, same classification, stored with your account email.
  • Email you send to any of our addresses (support@mark-it.co, hello@mark-it.co, tomer@mark-it.co and other addresses at mark-it.co, which are all forwarded by Cloudflare Email Routing into the same Gmail mailbox): your email address, display name, subject, message text and any attachments. AI tools we operate may read and sort new messages in that mailbox and suggest a reply for a person to review; routine support messages may receive a short automatic confirmation that the message was received (Section 5). A person reads every message. If you would rather your message be handled by a person only, say so in the email.
  • Searches you report as wrong. If you press "Not what you were looking for?" under a set of search results, we store the text of that one search, how many results it returned, and which part of the app you searched from, linked to your account so we can follow it up. Nothing is sent until you press the button, and we do not record the text of searches you do not report into our own database. Like the star feedback above, a reported search is also forwarded to our internal inbox at markit.ai.contact@gmail.com through Resend, so the search text and your account email reach that mailbox. Reports are kept separately from the star-rating feedback and are not sent to OpenAI for triage. They are deleted when you delete your account. Note that this is about our own database: the search page carries your terms in the page address, and the third-party analytics tools in Section 2.7 receive page addresses for every page you view.

2.9 Payment data (when paid tiers launch)

MarkIt is free and nothing is offered for sale, so we do not collect or process any payment data. We have never integrated a payment processor. If that ever changes, billing would be handled by a regulated payment processor and we would update this policy with that processor's name and role before charging anyone.

2.10 Early access requests

If you ask for early access through our request form (before you have an account), we collect:

  • The email address you submit, so we can reach you if you are approved.
  • A normalized (canonical) form of that email and its domain, derived and stored so we can de-duplicate repeat requests and prevent abuse of the form. For example, for Gmail addresses we strip dots and any "+tag" so the same mailbox is recognized as one requester.
  • Your free-text pitch (why you want in), so we can evaluate the request.
  • A one-way hashed form of your IP address for abuse prevention. We do not store your IP in raw form.
  • Your approximate country, derived from request headers at submit time.
  • The page you came from on our site.
  • Your browser user-agent string.

When you submit a request, the email and pitch plus the signals above are forwarded to an internal inbox at markit.ai.contact@gmail.com via Resend so our team can review it (the same way feedback is handled in Section 2.8). Please do not include sensitive personal information, or other people's personal data, in your pitch.

We use this only to evaluate and grant early access and to prevent abuse of the form. We do not name a separate sub-processor for this data; it is handled by the processors already listed in Section 4. Retention is described in Section 8.

2.11 API access and connected AI clients

You can generate a personal API key (it is prefixed mkt_) in Settings > Integrations and use it to connect an external AI client of your choice - for example Claude, ChatGPT, or Cursor - to your MarkIt account through our connector, an implementation of the open Model Context Protocol (MCP) at https://mark-it.co/api/mcp.

  • What we store for the key: a one-way SHA-256 hash of the key (new keys are stored only as this hash, never in plaintext), a short non-secret prefix and an optional name for display, and the time the key was last used. The key is shown to you once, at creation. Nothing about your saved content changes; this is only the credential record.
  • What a connected client can do: while a valid key is presented, the client can, at your direction, search and read your saved items, list your categories, read, create, reschedule, or cancel reminders on your items (cancelling deletes that one reminder; reminders synced to Google Calendar cannot be cancelled through the connector), and save new links and text notes into your library (each genuine new save counts toward your monthly capture limit; re-saving a link you already have is a no-op). It can also, only when you explicitly ask it to, send us a short feedback note - that note is handled exactly like feedback from the in-app widget (Section 2.8): stored with your account email, forwarded to our internal inbox, and auto-classified by OpenAI for triage; it is limited to a few notes per day. The client cannot delete your saved items, empty your library, or change your account settings. Requests are subject to anti-abuse rate limits.
  • Searching uses our AI sub-processors: when your client runs a search through the connector, your query - and, for ranking, short snippets of the matching items - is sent to OpenAI for embeddings and to Cloudflare Workers AI for reranking, the same providers (already listed in Section 4) that power in-app search. This is the one place the connector involves a MarkIt-side AI call.
  • Your chosen AI provider is your own tool, not our sub-processor: the AI client, and the provider behind it (for example Anthropic or OpenAI), are chosen and controlled by you. When your client reads your MarkIt data, that data is handled by that provider under its own terms and privacy policy, acting as a processor for you - not for us. For that reason your client's AI provider is not listed in our Section 4 sub-processor table. This is separate from the embeddings and reranking providers above, which are ours.
  • Two ways to connect, and to disconnect: you can connect with a personal API key (above), or - for connectors such as Claude.ai and ChatGPT that use OAuth - by authorizing MarkIt through a consent screen, which issues a short-lived access token bound to this service. Deleting your API key ends key-based access immediately, but does not end an OAuth connection. To end an OAuth connection, remove the MarkIt connector in your AI client (its tokens are short-lived and expire on their own), or email tomer@mark-it.co to have the grant revoked.

3. Why we process your data (purposes and legal bases)

EU and UK GDPR require us to disclose the legal basis for each purpose. The table below does that.

PurposeData categoriesLegal basisRetention
Create and run your account, save captures, run searchAccount, captured content, metadata, integration dataContract (Art. 6(1)(b))Until you delete your account
Answer email you send to our support and contact addresses, including sorting it and suggesting replies with AI tools we operateCommunication data (Section 2.8): your email address, name, subject, message text, attachmentsContract (Art. 6(1)(b)) where you are a user asking about the service; otherwise legitimate interest (Art. 6(1)(f)) in responding to people who write to usKept in our support mailbox for up to 24 months after the conversation ends, then deleted; earlier on request
Auto-categorize, tag, summarize, and embed your itemsCaptured content, derived metadataContract (Art. 6(1)(b)) - this is the product you signed up forUntil you delete the item or account
Deliver the reminders you scheduled - by email, WhatsApp, or Telegram - and the calendar events you addAccount data, reminder settings, your WhatsApp/Telegram link (if connected), Google tokens (if connected)Contract (Art. 6(1)(b))Until the reminder fires, then per normal retention
Engagement emails (product tips, milestone messages)Account data, engagement metadataLegitimate interest (GDPR Art. 6(1)(f)) combined with the soft-opt-in exemption in Member-State implementations of ePrivacy Directive 2002/58/EC Art. 13(2) (for UK users, Regulation 22(3) of the Privacy and Electronic Communications Regulations 2003), with one-click unsubscribe in every message. Where a Member State's implementation does not allow soft opt-in, we fall back to explicit prior consent.Until you unsubscribe
WhatsApp proactive messagesPhone hash, engagement metadataConsent (GDPR Art. 6(1)(a)) and the prior-consent requirement under ePrivacy Directive Art. 13(1) as implemented locally - explicit opt-in, default offUntil you reply STOP or revoke
Evaluate early-access requests + prevent form abuseEmail, pitch, hashed IP, country, source page, user-agentLegitimate interest (Art. 6(1)(f)); requester voluntarily submits30 days after a decision; 90 days if never decided
Fraud prevention, abuse monitoring, rate limitingAccount data, bot audit logs, heatmap session IDsLegitimate interest (Art. 6(1)(f))Server request logs retained per Vercel's platform policy; 30 days for bot audit; 500k-row circuit breaker for heatmaps
In-house product analytics (heatmaps, aggregate usage)Pseudonymous session data, page URLs, click and mouse-movement metadataLegitimate interest (Art. 6(1)(f)) - first-party, no cross-site tracking, input fields maskedRolling; ingestion pauses at 500k events
Understanding where new users get stuck (activation analytics)Page views with your user ID while signed in, campaign parameters, and product events (signup, save outcome, channel connected) with outcomes and reasons only - never your contentLegitimate interest (Art. 6(1)(f)) - first-party, no cross-site tracking, no advertising use, no content recordedPage views 365 days; product events 180 days; deleted automatically on a schedule
Page-speed monitoring (Core Web Vitals)Metric name and value, page path, pseudonymous session ID, country - not linked to your accountLegitimate interest (Art. 6(1)(f))365 days
Legal, tax, and regulatory complianceWhatever the law requiresLegal obligation (Art. 6(1)(c))As long as the obligation requires
Protecting the service (security incidents, disputes)All relevant dataLegitimate interest (Art. 6(1)(f))For the duration of the incident plus limitation periods

For direct marketing outside the soft-opt-in model (for example, a future newsletter to prospects who are not yet customers), we will rely on your explicit consent and will ask for it at the point of collection.

4. Third parties we share data with

We do not sell your personal data and we do not "share" it for cross-context behavioral advertising under the CCPA definition. The companies below process data on our behalf, as independent controllers where noted, under written contracts that restrict use to the purposes listed.

VendorRoleWhat they receiveHosting regionTransfer mechanism
Supabase, Inc.Processor (primary database, auth, storage)Everything stored in our database and the three storage buckets (items, user-uploads, ai-generated-images)Region tied to our Supabase project; email tomer@mark-it.co and we will confirm the current data-residency region before you rely on itStandard Contractual Clauses 2021/914 (EU Commission Implementing Decision 2021/914) incorporated into Supabase's DPA; Supabase is not DPF-certified
OpenAI, LLCProcessor (AI inference and embeddings)Item titles, URLs, scraped text, user notes, handwritten or document OCR text, and images for vision OCR. Output is not used to train OpenAI's public models per OpenAI's API data policy (default since March 2023). OpenAI retains API requests for up to 30 days for abuse monitoring under its default policy and we do not currently have a zero-data-retention flag on our account.USSCCs 2021/914; EU-US Data Privacy Framework certification where applicable
MarkIt (our own servers on Vercel)Controller (first-party fetch of public Instagram, Facebook, TikTok, LinkedIn, Threads and Pinterest posts, identifying as MarkItAgent, one fetch at save time, never logged in, never re-fetched)The public URL you saved. No user identifier is sent to Instagram.EU (Frankfurt)Same hosting and transfer terms as Vercel above
Apify Technologies s.r.o.Processor (social-media scrapers for TikTok, Facebook, LinkedIn, YouTube captions, and Instagram and Threads as a fallback when our own fetch fails)The public URL you saved. No user identifier.Czech Republic (EU)Intra-EU transfer at the Apify leg; data still transits US infrastructure upstream (Vercel/Supabase) under SCCs 2021/914
Proxy-Seller (SSV IT Provider Online Services Ltd)Processor (static residential/ISP proxy - relays our own server's one-time public-post fetch for Reddit and YouTube so the request originates from a residential IP; receives no user data and no cookies; active only when the proxy is configured)The public URL you saved, relayed to Reddit or YouTube. No user identifier, no cookies.Cyprus (EU); the relayed egress IP's country depends on the provisioned IP - email tomer@mark-it.co and we will confirm before you rely on itIntra-EU transfer at the Proxy-Seller leg; data still transits US infrastructure upstream (Vercel/Supabase) under SCCs 2021/914
Jina AI GmbH (r.jina.ai)Processor (reader extraction for non-social URLs)The public URL you saved. No user identifier.Germany (EU); requests to the r.jina.ai endpoint may be routed via global CDN for latency, but data is processed in the EUIntra-EU at the Jina leg; SCCs 2021/914 for any US CDN leg upstream
weserv.nl (operated from the Netherlands)Processor (on-the-fly image proxy and resize for social-CDN images)The source image URLNetherlands (EU); fetches public images through a CDNIntra-EU transfer at the weserv leg; standard SCC mechanism for any CDN leg that touches the US
Resend, Inc.Processor (email delivery)Your email address, email subject, full HTML body of messages we send you, including automatic acknowledgements sent to email you send to support@mark-it.coUSSCCs 2021/914
Anthropic, PBCProcessor (AI model used by tools we operate to sort email sent to our support and contact addresses and suggest replies)The email you send to any mark-it.co address: your email address, display name, subject, message text. Anthropic does not train its models on API inputs or outputs under its commercial terms.USSCCs 2021/914
Google LLC (Gmail)Processor (hosts the mailbox that receives email sent to our mark-it.co addresses; Cloudflare, Inc. forwards it there under Cloudflare Email Routing)The email you send to any mark-it.co address, in full, and our repliesUS (Google) and Cloudflare's global edge network for the forwarding legEU-US Data Privacy Framework (Google and Cloudflare are both DPF-certified); SCCs 2021/914 as a secondary mechanism
WaSender (WasenderAPI)Processor (WhatsApp Business API relay)Your phone number (without + prefix) and the bot message bodyRegion not published; treat as globalSCCs where applicable
Meta Platforms, Inc. (WhatsApp Business Platform)Independent controller for the messaging service; processor for message content per WhatsApp Business Data Processing TermsPhone number and message content, via WaSenderUS / IrelandEU-US Data Privacy Framework; WhatsApp Business Data Transfer Addendum
Telegram Messenger Inc.Independent controller for the messaging serviceTelegram user ID, chat ID, message contentGlobal (Telegram does not publish region pins)Reliance on Telegram's own terms
Microsoft Ireland Operations Limited, transferring to Microsoft Corporation (Clarity analytics)Independent controller. Microsoft's Clarity documentation states that Clarity is "GDPR-compliant as a data controller", so unlike the other entries in this table it is not acting solely under our instructionsSession recordings of pages you view, including pointer movement, clicks, scrolling and on-screen content; IP address; device and browser dataUS (your browser sends directly to clarity.ms; the Irish entity is the contracting party, not a hop in the data path)Standard Contractual Clauses (EU Commission-approved), Microsoft Ireland Operations Limited to Microsoft Corporation
Google LLC (Google Analytics 4)Processor for page-view and event analyticsPage views, interaction events, IP address, device and browser dataUSEU-US Data Privacy Framework
Google LLC (OAuth sign-in)Independent controller for your Google account; processor for our OAuth requestYour Google email, name, avatarUSEU-US Data Privacy Framework
Google LLC (Calendar API)Independent controller for your calendar dataEvent summary, description, start/end, reminder settings, and a private extended property containing the MarkIt item ID, for events you explicitly scheduledUSEU-US Data Privacy Framework
Vercel, Inc.Processor (application hosting and edge)All HTTP traffic to the Service; standard platform logs per Vercel's own policyUS default unless we pin a regionEU-US Data Privacy Framework (Vercel is DPF-certified); SCCs 2021/914 as a secondary mechanism
Cloudflare, Inc.Processor (bot-search relevance reranker via Cloudflare Workers AI)For each bot search call: the search query plus title and content snippets (up to 600 characters each) of candidate items from your library. Per Cloudflare's Workers AI Data Usage policy, Cloudflare does not use this content to train AI models, does not retain it beyond inference, and does not share it with other Cloudflare customers.Cloudflare global edge networkEU-US Data Privacy Framework (Cloudflare is DPF-certified); SCCs 2021/914 as a secondary mechanism
Cloudflare, Inc. (R2 object storage)Processor (encrypted off-site backup storage via Cloudflare R2)Encrypted backup copies of your account data: a nightly database dump (your items, categories, tags, notes, links, reminders, and account records) and a weekly copy of our storage buckets (uploaded images and documents). Backups are encrypted with age before they leave our systems, so Cloudflare stores only opaque encrypted blobs it cannot read; we hold the decryption key. See Section 8 for retention and post-deletion handling.Western Europe (EU)EU-US Data Privacy Framework (Cloudflare is DPF-certified); SCCs 2021/914 as a secondary mechanism

We keep written data-processing agreements (Article 28 GDPR) with every processor above. We maintain an internal record of sub-processors and will update this list before adding a new one that materially changes the data flow.

Beyond this list, we may disclose data where required by law, by court order, or by a government request with legal authority. We may also transfer data in a merger, acquisition, or asset sale; if that happens, we will notify you in advance and let you delete your account before the transfer takes effect.

Bot integrations

MarkIt offers two optional messaging-bot integrations: a WhatsApp bot delivered through WaSender (which relays the WhatsApp Business API) and a Telegram bot built on the official Telegram Bot API. You activate either integration explicitly from Settings > Integrations by clicking Connect. This section is the Article 13 notice for the data processing that begins at that click.

Lawful basis. Once you click Connect, MarkIt processes the content you send the bot under GDPR Article 6(1)(b) - performance of a contract. The contract is the user-facing agreement to provide bot capture, search, and reminder features in exchange for processing the messages and media you forward. Connect is the activation moment; disconnect is the termination of that sub-contract.

What we process and why.

  • Bot transport: WaSender (WhatsApp) or Telegram Messenger Inc. (Telegram) - relays the message between you and our servers. See the sub-processor table in Section 4 for hosting region and transfer mechanism.
  • AI classification, tagging, and summarization: OpenAI processes the text you forward (and OCR text from images and documents) so the bot can categorize, tag, summarize, and search your items. See Section 5 for full AI processing detail.
  • Social-media fetching: when you forward a public Instagram, Facebook, TikTok, LinkedIn, Threads or Pinterest URL, our own servers fetch the public post once (identifying as MarkItAgent, never logged in). As a fallback for Instagram, Facebook, TikTok, LinkedIn, Threads and YouTube, Apify fetches the public content. X (Twitter) posts are read by our own servers from X's public syndication endpoint, with the open-source fxtwitter and vxtwitter embed services - which receive only the public tweet address, never anything about you - as the fallback. Reddit posts (from Reddit's public RSS and oEmbed endpoints) and YouTube videos (from YouTube's public player and caption APIs) are read by our own servers, routed through a third-party proxy provider (Proxy-Seller) so the request comes from a residential IP - the proxy relays only the public post address, no user data and no cookies - with Apify as the fallback. See Section 4.
  • URL content extraction: when you forward any other public URL, our own servers fetch the public page once (identifying as MarkItAgent, never logged in) and extract the readable article text. Jina Reader (r.jina.ai) is the fallback when our own extraction returns too little. When both our own fetch and Jina fail (some sites block automated fetches), Apify (already a listed sub-processor) fetches the public page as a last resort - it receives only the public page address, no user data and no cookies. See Section 4.
  • Transactional email: Resend delivers any related notifications we send by email (for example, a reminder you scheduled through the bot). See Section 4.

We do not name sub-processors inline in this subsection because the canonical list lives in Section 4. If we add or replace a bot-related sub-processor we will update that table and bump the "Last Updated" date at the top of this page.

Retention of bot data. Bot audit logs follow the schedule documented in Section 2.4 and Section 8: whatsapp_messages auto-purge after about 30 days and are deleted immediately when you unlink WhatsApp; telegram_messagesare retained for abuse prevention pending a parallel delete-on-unlink path. Saved items captured through the bot are stored under your account and deleted only when you delete the item or your account.

Contract-activation record. To meet the Article 5(2) accountability requirement we keep a small bot_link_history record of when each integration was activated and disconnected, alongside the version of the consent notice shown to you at the time. This record survives disconnect for the limitation period and is used only to demonstrate, on request, that the integration was activated by you and when. It does not contain message content.

How to withdraw. You can end the integration any time. From the web, open Settings and tap Disconnect on the WhatsApp or Telegram card. From inside the bot, send /unlink on either platform. Disconnect terminates the sub-contract and removes your bot link record. Saved items captured during the integration are kept under your account until you delete them.

5. AI processing

We use OpenAI (GPT-5 nano for text tasks and embeddings, GPT-5 vision for image OCR, plus an image-generation model for the optional AI image feature) for the following:

  • OCR on images and screenshots you upload.
  • Categorization that picks which workspace category each new item belongs to, or leaves it in "Unsorted".
  • Tag generation (3 to 5 tags per item).
  • Summaries and AI-written headlines shown in the item modal.
  • Embeddings that power semantic search.
  • Two-agent bot search that ranks your saved items when you ask the WhatsApp or Telegram bot a question.
  • Engagement segmentation, run daily from your first-party activity (not from OpenAI) to decide whether a given email template is a good fit.
  • AI image generation, only when you explicitly trigger it and spend credits on it.

None of these decisions produce a legal or similarly significant effect on you under GDPR Article 22. Every AI label is reversible with one click in the UI: you can rename categories, edit tags, correct summaries, move an item back to "Unsorted", or delete the item entirely.

OpenAI processes your data as our processor under the OpenAI API data policy. OpenAI does not train its publicly available models on API data by default. OpenAI retains API requests for up to 30 days for abuse monitoring. We do not have an OpenAI zero-data-retention agreement on our account today.

We do not send Google Calendar data to OpenAI today. We do not send payment data to OpenAI today. We will update this policy and obtain your consent before adding any feature that would change this.

When you send feedback through the in-app feedback widget - or ask a connected AI client to send us a feedback note (Section 2.11) - the message text is sent to OpenAI (gpt-5.4-nano) so we can automatically classify it (category and priority) for our internal triage. OpenAI processes it on our behalf as described above and does not train its publicly available models on it by default.

Support email. Email you send to support@mark-it.co, hello@mark-it.co, tomer@mark-it.co or any other mark-it.co address may be read and sorted with the help of AI tools we operate on Anthropic's Claude models (Section 4), so that a person can reply faster. Those tools may suggest a reply for a person to review, and routine support messages may receive a short automatic confirmation that the message was received. Such confirmations are marked as automatic and use only pre-written text. A person reads every message. If you would rather your message be handled by a person only, say so in the email. None of this produces a legal or similarly significant effect on you under GDPR Article 22. Anthropic processes this data as our processor under its commercial terms and does not train its models on it.

6. Google Calendar integration (Limited Use disclosure)

If you connect Google Calendar, MarkIt requests the narrowest scope that fits the reminder feature: https://www.googleapis.com/auth/calendar.events, plus openid, userinfo.email, and userinfo.profile so we can identify your account.

MarkIt's use and transfer of information received from Google APIs adheres to the Google API Services User Data Policy, including the Limited Use requirements.

The Google API Services User Data Policy permits the use and transfer of information received from Google APIs only for the following purposes. The four exceptions below are quoted verbatim from Google's policy:

  1. "To provide or improve your appropriate access or user-facing features that are visible and prominent in the requesting application's user interface and only with the user's consent".
  2. "For security purposes (for example, investigating abuse)".
  3. "To comply with applicable laws".
  4. "As part of a merger, acquisition, or sale of assets of the developer after obtaining explicit prior consent from the user".

Applied to MarkIt: we use Google Calendar data only to create, read, update, and delete the reminder events you schedule inside MarkIt, which is a user-facing feature prominent in the MarkIt interface. We do not use Google Calendar data to serve advertisements, including retargeting, personalized, or interest-based advertising. Humans at MarkIt do not read Google Calendar data except where an exception above applies. Beyond those exceptions, MarkIt voluntarily commits that if a merger, acquisition, or asset sale ever occurs, we will give you advance notice and the opportunity to delete your data.

We do not use Google Calendar data to train AI or machine-learning models of any kind, and we do not send Calendar data to OpenAI.

OAuth tokens live only in your browser's NextAuth session cookie. We do not copy them to our database. To revoke access at any time, sign out of your Google account in MarkIt or visit myaccount.google.com/permissions and remove MarkIt. There is no "revoke" button inside MarkIt yet - this is on our roadmap.

7. International transfers

MarkIt is operated from Israel. Our primary database (Supabase) and application hosting (Vercel) are in the United States. Several sub-processors are located in the US, the EU, or operate globally. Israel is currently recognized by the European Commission as offering an adequate level of data protection for transfers from the EEA under Commission Decision 2011/61/EU. If that decision is amended or revoked, we will update this policy and apply SCCs to those transfers.

Your data is transferred outside the European Economic Area, the United Kingdom, and Israel when it reaches US-based sub-processors. We rely on two mechanisms, in this order of priority for each recipient: (a) the EU-US Data Privacy Framework and its UK Extension, where the recipient is certified (currently Vercel; see dataprivacyframework.gov/list for current status); and (b) the European Commission's Standard Contractual Clauses (Implementing Decision (EU) 2021/914) together with the UK Addendum to the SCCs, which are incorporated into each sub-processor's Data Processing Agreement. For Israel, we rely on the EU adequacy decision for Israel. For each sub-processor we assess the level of protection in the receiving country before transferring. You can request a copy of the transfer-impact assessment from tomer@mark-it.co.

For transfers from Israel, where the recipient country is not on the Israeli Privacy Protection Authority's list of adequate jurisdictions, we use a written contract that binds the recipient to standards equivalent to the PPL and the Israeli Data Security Regulations 2017.

8. Data retention

  • Account data: kept until you delete your account. On deletion we hard-delete the content you saved (items, notes, tags, categories, reminders) and your account record immediately, along with embeddings, credits, milestones, referrals, storage files in all three buckets, and used invite codes. Bot audit logs (WhatsApp and Telegram message metadata) and engagement-message logs currently persist as orphaned records with your user ID removed, and are purged when our retention cleanup runs. We are tightening this so deletion cascades everywhere; until then, if you want your bot audit logs deleted immediately, email tomer@mark-it.co.
  • Backups: we keep encrypted backups of your data to protect against accidental loss. Database backups are retained up to 30 days; image and file backups up to 60 days. All backups are encrypted at rest and stored in the EU (Cloudflare R2, EU jurisdiction). Supabase additionally retains its own platform backups per its policy. When you delete your account, your data is removed from live systems immediately. Backup copies expire and are deleted within 60 days. In the unlikely event we need to restore from a backup that pre-dates your deletion, we maintain a separate deletion log (containing only one-way SHA-256 hashes of your email and phone, never the raw values) and re-apply pending deletions to the restored data so it does not reappear.
  • Captured content: kept until you delete the item or the account.
  • Bot audit logs (whatsapp_messages, telegram_messages): whatsapp_messages are set to auto-purge after about 30 days and are deleted immediately when you unlink WhatsApp. telegram_messages are currently retained for abuse prevention and are not yet deleted when you unlink Telegram - we will add a parallel delete on the Telegram side; in the meantime, email tomer@mark-it.co to have your Telegram audit rows removed.
  • Engagement message records: retained for the life of the account for compliance with direct-marketing record-keeping rules and to honor unsubscribe requests.
  • Heatmap events: stored pseudonymously by session ID. Ingestion pauses automatically when the table reaches 500,000 rows; older events are rotated out when we compact.
  • Server logs at Vercel: retained per Vercel's platform policy (typically several days to a few weeks, depending on plan).
  • OAuth tokens: Google Calendar tokens live in your NextAuth session cookie only. They are never persisted on our servers and expire per Google's token lifetime. Revoking access through Google immediately invalidates them.
  • AI request metadata at OpenAI: up to 30 days under OpenAI's default abuse-monitoring retention, then deleted by OpenAI. We do not copy that data.
  • Suppression list for unsubscribed emails: kept indefinitely to honor your opt-out.
  • Email you send to our support and contact addresses: kept in our support mailbox for up to 24 months after the conversation ends, then deleted; deleted earlier if you ask. Automatic acknowledgements we send are recorded in a send log (recipient address, subject, time, which acknowledgement text was used) kept for 12 months so we can honor rate limits and audit what was sent.
  • Early access requests: request records (email, pitch, hashed IP, country, source page, user-agent) are deleted within about 30 days of a decision (approve or deny). Requests that are never decided are deleted within about 90 days.

9. Security

  • HTTPS and TLS 1.2+ enforced at the edge by Vercel.
  • Security headers set by the application: X-Frame-Options, X-Content-Type-Options: nosniff, Referrer-Policy: strict-origin-when-cross-origin, and a Permissions-Policy that disables camera, microphone, and geolocation. A Content Security Policy is in place in report-only mode while we tune it; we will move it to enforcing once tuning is complete.
  • Row-Level Security is enabled on the tables that store user content, bot links, engagement messages, and audit logs. Every query in application code is scoped to the authenticated user.
  • Authentication uses Supabase Auth (industry-standard password hashing, currently bcrypt per Supabase's documentation) and JWT-based session cookies. Google OAuth sign-in is supported.
  • Rate limiting on heatmap, WhatsApp, and send retries, plus an atomic daily-message cap per user for the bots.
  • LLM prompt-injection hardening on every AI system prompt, including a sanitizer for user-supplied text before it reaches the model.
  • An admin audit log records staff actions that touch user data.
  • Account deletion requires you to type your email address as a confirmation challenge.

No system is perfectly secure. If a breach is likely to risk your rights and freedoms, we will notify the relevant supervisory authority within 72 hours of becoming aware of it (GDPR Art. 33). Where the risk to you is high, we will notify you without undue delay (GDPR Art. 34).

10. Your rights

You have the following rights regardless of where you live. We will honor them within 30 days, extendable once by an additional 30 days for complex requests (GDPR allows up to three months; we target 30 days as our standard).

  • Access: ask for a copy of the personal data we hold about you.
  • Rectification: correct data that is wrong.
  • Erasure: delete your account and the data tied to it. You can do this yourself from Settings > Delete Account.
  • Restriction: ask us to pause processing while we investigate an accuracy or objection request.
  • Portability: receive your data in a structured, machine-readable format. We do not currently have a self-service export button. On written request to tomer@mark-it.co we will prepare an export, typically as JSON with attached media links, within 30 days.
  • Objection: object to processing based on legitimate interests (including engagement emails). We will stop unless we can show compelling grounds.
  • Withdraw consent: where we rely on consent (for example, WhatsApp proactive messages), you can withdraw it at any time without affecting past processing.
  • Not be subject to solely automated decisions with legal or similarly significant effect: MarkIt does not make such decisions about you. Our AI produces labels you can change.
  • Complain to a supervisory authority: see section 14.

To exercise any of these rights, email tomer@mark-it.co from the address on your account or from an address you can verify you control. We may ask you to confirm your identity if the request is not obviously from you.

Channel-specific opt-outs already built in

  • Email marketing: one-click unsubscribe in every engagement email. We implement RFC 8058 (List-Unsubscribe and List-Unsubscribe-Post headers) so Gmail, Yahoo, and other clients can unsubscribe you without round-tripping through our site.
  • WhatsApp proactive messages: to opt out, reply STOP in English. (We are adding Hebrew STOP - reply STOP in English for now; or unlink WhatsApp from Settings at any time to stop all WhatsApp messages.)
  • Telegram proactive messages: to stop Telegram bot messages, use /unlink in the bot (links to your MarkIt account are removed) or unlink Telegram from Settings. Your Telegram audit logs are retained per the general retention schedule and are removed on account deletion.
  • Google Calendar: revoke access at myaccount.google.com/permissions.

11. California residents (CCPA / CPRA)

This section applies if you are a California resident. We treat these rights as extending to all US residents regardless of state.

Categories of personal information we collected in the preceding 12 months (using the CCPA's statutory categories under Cal. Civ. Code § 1798.140(v)(1)):

CCPA categoryCollected?Source
A. Identifiers (name, email, IP)YesDirect from you / Google OAuth
B. Customer records (CA CC 1798.80(e))No-
C. Protected classificationsNo-
D. Commercial informationYes (credit/feature usage)Inferred from activity
E. Biometric informationNo-
F. Internet or network activityYes (heatmap events: clicks, scroll, mouse movement)In-browser telemetry
G. Geolocation dataNo (only coarse IP region for rate limits, not stored)-
H. Sensory data (audio, visual, thermal, olfactory)No-
I. Professional or employment informationNo-
J. Education informationNo-
K. InferencesYes (category/tag/engagement-segment assignments)Derived by our AI

Sources: directly from you, from Google (OAuth profile), from Apify/Jina (public content you chose to save), and from your device (session data).

Purposes: see section 3. Business purposes: providing the service, preventing fraud, and ensuring security.

We do not sell personal information. Disclosures to the vendors in Section 4 are made to service providers under written contracts that meet CCPA § 1798.140(ag) requirements and that restrict use of the personal information to providing the services to MarkIt. We do not "share" personal information for cross-context behavioral advertising under the CCPA definition. We honor the Global Privacy Control signal as a valid opt-out of sale or sharing, even though we have no such activities.

Sensitive Personal Information: we do not collect any SPI (CA CC § 1798.140(ae)) - no government IDs, financial account numbers, precise geolocation, racial/ethnic/religious origin, union membership, genetic/biometric data, health information, or sexual-orientation data. Users may upload such data into their own captured content (notes, documents); they remain responsible for that, and we treat it as User Content subject to standard protections, not as SPI we collect for our own purposes.

Shine the Light (Cal. Civ. Code § 1798.83): we do not disclose personal information to third parties for their own direct-marketing purposes. California residents are entitled to request this information once per year; the response for MarkIt would be: none.

Your California rights: to know, to delete, to correct, to opt out of sale/share (not applicable to us), to limit use of sensitive PI (not applicable, since we collect no SPI), and not to be retaliated against for exercising these rights.

Automated decision-making technology (ADMT): our AI categorization, tagging, summarization, and engagement segmentation are ADMT under the CPPA definition. California's Automated Decisionmaking Technology regulations were adopted by the California Privacy Protection Agency in September 2025 and are being implemented on a phased schedule. MarkIt's AI features do not make decisions that produce legal or similarly significant effects on users (our AI only suggests a category, generates a tag, summarizes text, or ranks search results - all of which you can override), so the ADMT transparency obligations that apply to us are limited. We will update this policy if any MarkIt feature begins producing ADMT-covered decisions.

How to exercise: email tomer@mark-it.co. We verify by replying to the address on file.

12. UK residents (UK GDPR / DPA 2018)

If you are in the UK, the rights listed in section 10 apply under the UK GDPR. You can complain to the Information Commissioner's Office at ico.org.uk/make-a-complaint. We rely on the "soft opt-in" in Regulation 22(3) of the Privacy and Electronic Communications (EC Directive) Regulations 2003 for engagement emails to existing users, and every such message carries a one-click unsubscribe. We will name a UK Article 27 representative in this policy before the Service is generally available to UK residents outside the invite-only early access.

13. Israel residents (PPL 5741-1981, as amended by Amendment 13, 14 August 2025)

MarkIt is operated from Israel and the controller is an Israeli sole proprietor. The Israeli Privacy Protection Authority is our home supervisory authority.

  • Access and correction: on request to tomer@mark-it.co, we will provide the data we hold about you in Hebrew, English, or Arabic and correct inaccuracies within 30 days.
  • Database registration: our database is not registrable under the post-Amendment-13 thresholds (we do not collect personal data to transfer to others as a business activity, and we are below the sensitive-data thresholds).
  • Data Protection Officer / Privacy Supervisor: we have not appointed one under Amendment 13 thresholds. Privacy questions go to tomer@mark-it.co.
  • Hebrew translation: the authoritative version of this policy is in English. A Hebrew translation is available on request. We plan to publish a bilingual /privacy page; until then, email us.
  • Cross-border transfers: see section 7.

14. Supervisory authorities

You can complain to the supervisory authority where you live, where you work, or where you believe a violation occurred.

We would appreciate the chance to address your concerns before you contact a regulator. Email tomer@mark-it.co first.

15. Children

MarkIt is not directed to children. You must be at least 13 years old to use MarkIt in the United States (COPPA), at least 16 in the European Union unless your member state has lowered the digital-consent age to 13, 14, or 15, and at least 18 in jurisdictions where that is the default age of contractual capacity unless your legal guardian has agreed on your behalf.

We do not have a technical age gate today. We rely on your acceptance of the Terms and on user reports. If you believe a child under the applicable minimum age has used MarkIt, email tomer@mark-it.co and we will delete the account.

16. Cookies and local storage

Our own cookies and browser storage are strictly necessary or first-party UI state, and we set no advertising or cross-site tracking cookies of our own. Since September 1, 2026, however, the two third-party analytics tools in Section 2.7 (Microsoft Clarity and Google Analytics) set their own cookies in your browser, so this page can no longer say that only strictly necessary cookies are used.

NameTypePurposeLifetime
_clck, _clsk, CLID (Microsoft Clarity)Third-party cookies, set by MicrosoftSession recording and analytics (Section 2.7). Not set by us and not strictly necessaryPer Microsoft's own retention for these cookies
_ga, _ga_* (Google Analytics)Third-party cookies, set by GooglePage-view and traffic analytics (Section 2.7). Not set by us and not strictly necessaryPer Google's own retention for these cookies
sb-* (Supabase Auth)Cookie (HttpOnly)Keeps you signed inUntil you sign out or the session expires
NextAuth session cookieCookie (HttpOnly)Holds your Google Calendar OAuth token when you connect CalendarUntil you sign out or the session expires
markit_invite_codeCookie (SameSite=Lax)Carries your invite code across the Google OAuth redirect during sign-up10 minutes
mk_sidsessionStoragePseudonymous per-tab session ID, shared by heatmap analytics, page views, and page-speed measurementsCleared when you close the tab
markit_has_items, markit_tz, and other UI flagslocalStorageRemember sidebar state, timezone, seen/dismissed prompts, and recent search queriesUntil you clear site data

This policy previously said that we showed no cookie consent banner because only strictly necessary cookies were used, and undertook to present a banner with an equal-weight "Reject All" option before any non-essential technology loaded. Third-party analytics were added on September 1, 2026 and that banner has not been built, so the tools currently load on page load without asking you first. We are stating that plainly rather than removing the undertaking. If you would prefer not to be recorded while we resolve this, email tomer@mark-it.coand we will exclude your account.

17. Who at MarkIt can see your data

MarkIt is a small team. Access to production data is limited to the controller (Tomer Ezri) and staff on the ADMIN_EMAILS allow-list, and is logged in an admin audit table. Staff access is used only to investigate abuse, diagnose a bug you reported, respond to a legal request, or build a feature you asked for. We do not read your content for marketing purposes. Email you send to our support and contact addresses may also be read by the AI tools described in Section 5, which we operate and which act on our instructions; they have read-only access to that mailbox and no access to your saved content or the production database.

18. Direct marketing (email and messaging)

  • Email: we send product engagement emails to existing users under the ePrivacy "soft opt-in" (Directive 2002/58/EC Article 13(2), as transposed in each EU member state, and PECR Regulation 22(3) in the UK), which permits marketing emails to existing customers about similar products or services where: (a) we collected your email during a sale or sign-up, (b) every email includes a one-click unsubscribe, and (c) you did not opt out when we first collected your email. Our lawful basis under GDPR Art. 6(1)(f) is our legitimate interest in keeping active users informed about MarkIt features, balanced against your rights; you can object at any time via the unsubscribe link. For US recipients we comply with CAN-SPAM. Every message identifies us as the sender, uses an accurate subject line, and includes a one-click RFC 8058 unsubscribe header plus a visible unsubscribe link. Our postal address for CAN-SPAM purposes is: Helsinki St 4, Tel Aviv 6299604, Israel.
  • The bi-weekly recap: a summary of what you saved over the previous 14 days. Email is its default channel, and it is covered by the soft opt-in described above rather than by a separate consent. On 13 August 2026 we enabled it by default for existing accounts, so you may receive it without having switched it on yourself; accounts that had already unsubscribed from our email were left untouched. We skip it entirely in any fortnight where you saved fewer than two things. Every recap carries a one-click unsubscribe and a link to Settings, where you can change its channel or switch it off without affecting any other email.
  • WhatsApp and Telegram: proactive messages (the bi-weekly recap, occasional re-engagement nudges) go out only to users who have explicitly chosen a chat channel. Receiving the recap on WhatsApp or Telegram is a Pro feature and is never the default; email is the default for everyone. To pause future bot messages, send /mute to the bot at any time, or toggle "Recap channel" in Settings. /mute stops both recaps and re-engagement nudges; you can resume with /unmute. The legacy STOP keyword (in English) is still honoured immediately, and /unlink disconnects the bot from your account entirely. Both keywords take effect on the next inbound message.
  • Frequency caps: at most one engagement message per user per day across channels, at most two WhatsApp proactive messages per month, a 24-hour gap between channels, progressive reduction after two ignored emails or three ignored WhatsApps, and a 90-day sunset after zero engagement.

19. Changes to this policy

We will post a new "Last Updated" date at the top of this page for every change. For material changes we will also email the address on your account and show a banner in the app for at least 30 days before the change takes effect. Corrections and additions that expand your rights may take effect immediately. If you do not agree to a material change, you can delete your account before the effective date.

20. Contact