17:00Legal

Privacy Policy

Last updated:November 14, 2026

Overview

Loqalit is a Chrome extension for localization QA professionals that scans web pages and analyzes translation quality using AI. This privacy policy describes how the Extension collects, uses, and protects your information. It also covers Loqalit Translate (translate.loqalit.com), our website translation service, the Loqalit platform (app.loqalit.com), the Loqalit connector for AI assistants (such as Claude and ChatGPT), and our website (www.loqalit.com), each in its own section below.

Information We Collect

1. Page Content (During Scans)

When you click "Scan Page," we extract text content from the active webpage:

  • Text content (headings, paragraphs, buttons, form labels, etc.)
  • Optional language and text-direction attributes for each text block
  • A minimal source-type hint when text comes from a non-default source (for example an ARIA attribute or a CSS pseudo-element)

We intentionally do not collect DOM locators such as CSS selectors or element positions — scan payloads contain text content and the metadata above only.

Storage: Scan data is stored locally in your browser (IndexedDB for history, chrome.storage.session for active scans). Data is sent to our servers in exactly two cases: when you run the AI analysis feature, and — for signed-in users who have opted in — through the scan-history sync described below.

Purpose: To display scan results, perform quality checks, enable highlighting, and persist scan history locally.

2. Scan-History Sync (Opt-In, Signed-In Users)

If you are signed in and explicitly enable scan-history sync (one-time dashboard prompt, changeable later in Settings → Account), the dashboard uploads your local scan history to your Loqalit workspace:

  • Page URLs of scanned pages
  • Scan timestamps
  • Block and word counts
  • Quality scores

Sync is off until you enable it; no scan history is uploaded before that. Disabling sync stops future uploads but does not delete entries already uploaded — clearing local data (Reset) or uninstalling the extension also does not delete server-side copies. You can request deletion of synced data at any time (see Contact below).

When a signed-in scan is associated with a workspace or project, its captured workspace/project context may also be stored with the analysis so the dashboard can distinguish where the scan originated from its later project assignment.

3. Authentication Data (AI Features)

If you use AI analysis with a signed-in account:

  • You sign in via Clerk authentication through a browser popup window
  • Your email and user ID are stored on our backend to manage access and usage quotas
  • Session state is managed via cookies on the auth host domain — no tokens are stored locally

4. Workspace, Invitation, and Billing Data

For signed-in workspace features, our backend stores the data needed to provide the selected workspace and its access controls:

  • Workspace, team, and project names and identifiers; membership roles; and

each user's selected workspace/project context

  • Invitation recipient email, role, status, expiry, inviter, and an opaque

invitation reference used by the invitation landing page

  • Billing plan and subscription status, seat/usage records, invoices, and

checkout/portal/cancellation intent metadata such as action, plan, provider reference, status, and timestamps

Hosted payment details are entered directly with Dodo Payments. Loqalit does not receive or store full card details. Opening checkout, a portal, an email draft, or a return URL is not treated as proof that payment or delivery completed; verified provider/backend state remains authoritative.

5. Usage Data

Our backend logs AI analysis requests including:

  • Timestamp and response time
  • Page URL and page domain
  • Block count, character count, and token usage
  • Model used, cost, and error information (if any)
  • Finding count from AI analysis

For signed-in analyses, the backend also stores the finding contents returned by the AI (message, rationale, and suggestion text — which may quote text from the analyzed page) in our backend database so your analysis results are available in the dashboard. It may also store the immutable originating workspace/project context separately from the analysis's current project assignment. When R2 storage is configured, the backend may additionally store the full request payload including system/user prompts and raw text blocks.

This helps us monitor service health, enforce usage limits, and improve analysis quality.

6. Guest Analysis Data

Guest users (not signed in) can use AI analysis with limited quota:

  • Identified by a locally-generated UUID device ID
  • Same request data is logged as authenticated users, with device ID instead of user ID
  • Default quota: 1 analysis per device

7. Crowdin Integration (Optional, User-Configured)

If you configure the optional Crowdin integration with your own Crowdin Personal Access Token:

  • The token is stored as authentication information in chrome.storage.local on your device and is sent only to Crowdin's API (api.crowdin.com) to act on your behalf
  • Access to api.crowdin.com is an optional browser permission, requested only when you set up the integration
  • When you explicitly push a finding to Crowdin, the Extension sends the AI finding text, the page URL and title, and a short excerpt of the rendered page text (up to ~280 characters) to your own Crowdin project as an issue
  • When screenshot evidence is included with a push, the Extension captures the visible tab and uploads the screenshot to your Crowdin project

Crowdin data goes directly from your browser to Crowdin — Loqalit servers do not receive or store your Crowdin token or pushed content.

7b. Phrase Strings Integration (Optional, User-Configured)

If you configure the optional Phrase Strings integration with your own Phrase API token:

  • The token is stored as authentication information in chrome.storage.local on your device and is sent only to Phrase's API (api.phrase.com) to act on your behalf
  • Access to api.phrase.com is an optional browser permission, requested only when you set up the integration
  • To find the key a finding belongs to, the Extension searches your Phrase project's translations for the rendered text of the affected string
  • When you explicitly push a finding to Phrase, the Extension adds a comment to the matching key in your own Phrase project containing the finding's severity and category, the page title and URL, the finding summary, rationale and suggestion, and a short excerpt of the rendered page text (up to ~280 characters). If you enable comment routing, a follow-up mention comment is added as well
  • When screenshot evidence is included with a push, the Extension captures the visible tab, uploads the screenshot to your Phrase project, and attaches a marker linking it to the key

Phrase data goes directly from your browser to Phrase — Loqalit servers do not receive or store your Phrase token or pushed content.

8. Product Diagnostics

The Extension can save operational diagnostic entries locally when expected product operations partially fail, such as skipped screenshot evidence capture. Backend upload is disabled by default and controlled by release configuration. When enabled, uploaded product diagnostics are limited to:

  • Diagnostic ID, event kind, source context, operation, severity, and stable

error code

  • Extension/app version, release channel, browser version, and environment
  • Page origin plus redacted path or hashes; raw query strings are excluded
  • Bounded counts such as captured/skipped targets and block count
  • Bounded per-target metadata such as block ID, finding index, category,

severity, and error code

Uploaded product diagnostics do not include screenshots, DOM snapshots, full page text, raw AI payloads, cookies, authorization headers, Clerk tokens, raw request headers, or raw query strings.

Vendor crash monitoring is currently disabled. If enabled later, it will be a separate exception-reporting lane for sanitized crashes only; session replay is not part of the current product.

9. User Preferences

Configuration preferences stored in Chrome storage:

  • Quality thresholds (MQM settings)
  • Scan configuration options
  • Onboarding completion state and distinct scan-history sync consent states

(unset, deferred, enabled, or explicitly declined)

Storage: Most preferences use chrome.storage.sync. Onboarding state, sync consent, and a few local-only UX flags use chrome.storage.local.

Loqalit Translate (Website Translation Service)

Loqalit Translate is a separate product: website owners add a small script to their site, and their visitors can read the site in other languages. Because it runs on our customers' websites, it handles data from two groups of people.

If You Are a Translate Customer (a website owner)

We collect your account information (through Clerk, as above), your site configuration (domain, chosen languages, switcher appearance, glossary rules, and translation style settings), an optional site key that identifies your site to our service, and billing information (through Dodo Payments, as above; we never receive card data). Text content from your website's pages and its translations are stored in your site's translation memory (see below) for as long as your site is registered. When you remove a site, its translation memory and configuration are deleted.

If You Are a Visitor to a Website That Uses Loqalit Translate

When you view a page on a site that has installed Loqalit Translate and choose a language:

  • Page text — the visible text of the page you are viewing is sent to our service so it can be translated. Text that already has a stored translation is served from the site's translation memory without further processing.
  • Page address — the URL of the page (without query parameters or fragments) is recorded in the site's page inventory so the site owner can see which pages are translated.
  • A random session identifier — generated in your browser for the current tab, not linked to any account or identity, used only to operate the translation service safely.
  • Your IP address — processed transiently and in hashed form for rate limiting and abuse prevention; it is not stored with any content.
  • Your language choice — saved in your own browser's local storage so the site stays in your language. The widget sets no cookies and contains no analytics or tracking.

Personalized content is deliberately kept out of storage. A piece of text is only written to a site's shared translation memory after it has been seen from two independent browsing sessions. Text that appears only in your session — your name, an order number, private account details — may be translated for your screen but is never stored.

Where Translation Data Lives and How It Is Processed

Translation memories (source text and translations) are stored at the network edge on Cloudflare infrastructure. When new text needs translating, it is sent from our service to an AI translation provider (see the third-party table) as text only, wrapped so that it is treated strictly as content to translate. Site owners' glossary rules and style settings are applied during translation.

The Landing-Page Demo and Site Previews

The demo on our landing page translates a page you point it at without storing anything: demo translations are shown to you and then discarded, and demo access expires automatically. Site previews (used during onboarding) fetch the customer's page the way a normal visitor would, forward no cookies in either direction, and are marked non-indexable for search engines.

Who Is Responsible for What

For visitor data processed on a customer's website, the website owner decides to use Loqalit Translate on their site and is responsible for informing their visitors; Loqalit processes this data on the owner's behalf, only to provide the translation service described here. If you are a visitor with questions or requests about a specific website's data, contact that website's owner; you can also reach us at info@loqalit.com.

Loqalit Platform (app.loqalit.com)

The Loqalit platform is the service your team uses at app.loqalit.com. You choose the pages of your product (and, where enabled for your workspace, your mobile apps) and the languages to check; Loqalit checks them on a schedule or when you ask, keeps the findings, and helps you fix them. Your workspace's administrators decide what is checked and who has access. For the pages, apps, and content you point us at, we process data on your behalf.

Scans of Your Pages

  • What we open: our scan workers open the pages you list, live or on staging, in a browser on our servers. They visit as an anonymous visitor unless you set up a sign-in (see below).
  • What we read: the visible text of each page, its language and layout, and a screenshot of the page.
  • Suggestions: to suggest pages you are not watching yet, the Scout agent may read your sitemap and the links on your pages. A suggested page is added only when someone in your workspace accepts it.
  • What we keep: for each finding, the page URL, the affected text, the finding itself (category, severity, explanation, and suggested fix), and a cropped screenshot of the area. Screenshots are deleted when your workspace's retention period has passed (90 days unless agreed otherwise); findings and their history are kept for the life of the account.
  • AI analysis: page text and screenshots are sent to our AI provider for analysis (see Third-Party Services).

Pages Behind a Login

If you set up a sign-in for scans, you give us a login page, a test account and, if that account uses an authenticator app, its two-factor seed. We store these encrypted and open them only inside the scan worker that signs in for a run. They are never shown again in the app or returned by any API. Please use a dedicated test account, not a person's own account. Our workers never solve CAPTCHAs.

Mobile Apps (Where Enabled)

  • Agent runs: from a store link you provide, our agent installs your app on an emulator we run, goes through its screens in each language you choose, and takes screenshots.
  • Agent sign-up: if you allow it, the agent signs up for your app with an email address of its own on qa.loqalit.com. We receive the confirmation emails sent to those addresses and keep only their subject, text, links, and one-time codes, for that sign-up.
  • Uploads from your CI: instead, your build pipeline can send us screenshots of your app with a project API key.
  • Retention: mobile screenshots and crops are deleted when the retention period has passed, as for web scans (90 days unless agreed otherwise).

The Assistant, Dictation, and Slack

  • Assistant: the assistant in app.loqalit.com answers questions about your projects and can propose changes for you to accept. Your question and the project data needed to answer it are sent to our AI provider. Your question, a record of the steps taken, and their cost are kept with the project.
  • Dictation: if you use the microphone button, the recording is sent through our server to our AI provider to be turned into text. We do not store the recording.
  • Slack: if a workspace administrator connects Slack, we store the Slack bot token encrypted. Anyone who can mention the Loqalit app in that Slack workspace can ask about the projects the administrator who connected it can see. Questions and answers pass through Slack and our AI provider.

Integrations, API Keys, and Webhooks

  • Crowdin and Phrase: if you connect them with a token, we store the token encrypted and never return it to any screen or API. When someone in your workspace accepts a fix, our servers find the matching string in your project and add an issue or comment with the finding, the page URL, and, where available, a screenshot.
  • API keys: we store only a hash of each key you create.
  • Release gate: the GitHub Actions template calls our API with your key to scan your staging pages and report the result to your pipeline.
  • Webhooks: they send signed event messages to the addresses you configure. A completed scan sends its status, languages, page count, and severity counts; an accepted finding sends the finding and the accepted fix, which may quote text from your pages.
  • Weekly report: if it is turned on, it is emailed every Monday to the people named in the project.

Where Platform Data Is Processed

Platform data is stored in our backend database and file storage. Scans run on servers and build runners we rent (see Third-Party Services), and analysis runs at our AI provider. These providers may process data outside the European Economic Area, including in the United States. We do not currently offer EU-only data storage.

Loqalit in AI Assistants (MCP Connector)

You can connect Loqalit to an AI assistant that supports the Model Context Protocol (MCP), such as Claude, ChatGPT, Cursor, or Codex. The assistant can then call Loqalit's tools on your behalf: list your projects, read their findings, start scans of public web pages, and check text you provide.

What We Receive

  • Sign-in: when you connect through an AI assistant, you sign in to Loqalit through Clerk. We receive your Clerk user ID and look up your email address to identify your account. If you have no Loqalit account yet, one is created at that moment. Developer tools can instead connect with an API key you create in the Loqalit app.
  • Tool inputs: only what the assistant sends to a Loqalit tool: the page URLs you ask us to scan, target languages, project or scan identifiers, and the text you ask us to check. We do not receive your conversation with the assistant, your other messages, or your files.
  • Client name: the name the assistant's software reports about itself (for example openai-mcp), used to count usage per assistant.

What We Do With It

  • Scans: we load the public pages you name in a browser on our servers and analyze their visible text the same way an extension analysis does. The resulting findings are stored in your Loqalit workspace, as described in Usage Data above.
  • Text checks: the strings you submit are sent to our AI provider for analysis (see Third-Party Services) and the findings are returned to the assistant. The strings themselves are not saved to your workspace; the request is logged with counts, token usage, and cost, as for other AI analyses.
  • Accounts created through an assistant: we record that the account was created through an AI assistant and which one, and grant a small one-time allowance of text checks. If the account's email address is on a company domain (not a personal mailbox provider), our team receives an internal notification with that email address, its domain, and the assistant's name, and may contact you about your Loqalit account.
  • Limits: we keep per-account and per-key request counters to enforce rate limits and plan allowances.

What the Assistant Receives

Tool results are sent to the AI assistant you connected: project names, findings (which may quote text from the analyzed pages), scan status, and links to the Loqalit app. Once delivered, that content is processed by the assistant's provider under its own privacy policy, not ours. Page text in results is labeled as untrusted content from third-party websites.

Your Controls

  • Disconnect Loqalit at any time in your assistant's settings, or revoke an API key in the Loqalit app (Settings → Webhooks & API).
  • Workspace admins can turn off AI assistant access for their workspace on the account page; assistants then receive nothing from that workspace.
  • Request deletion of your account and its data at any time (see Contact).

Our Website (www.loqalit.com)

This section covers visits to our website, www.loqalit.com. The sections above cover the Extension, your Loqalit account, Loqalit Translate, and the AI-assistant connector.

Hosting

The website is hosted on Cloudflare Pages. Cloudflare processes standard web request data, such as your IP address, browser, and the page requested, to deliver the site and protect it from abuse.

If you choose "Allow analytics" in the cookie banner, we load Google Analytics 4 to understand which pages are useful and how visitors find us. Until you do, or if you choose "Reject", the Google Analytics script is not loaded and no analytics cookies are set.

  • Data collected: pages viewed, the site that referred you, approximate location (derived by Google), device and browser type, and interactions such as clicks on links and buttons. Google Analytics 4 does not log or store IP addresses.
  • Cookies: _ga and _ga_<ID>, which tell one visitor from another.
  • Retention: Google Analytics data is kept for 14 months.
  • Processor and transfers: Google processes this data on our behalf and may process it outside the European Economic Area, including in the United States.

You can change your choice at any time with "Cookie settings" at the bottom of every page. If you withdraw consent, Google Analytics stops loading and its cookies are removed. Your choice is saved in your browser's local storage so we don't ask again.

Free QA Report and Demo Requests

When you send the free QA report form or the demo request form, we receive what you enter: your work email, your company website and, depending on the form, your name, the languages and pages you name, what you want Loqalit to check, how you check translations today, and any notes.

With the request we also receive two details about how your visit started: the first page you opened on our website and the site that referred you (for example a search engine or an AI assistant), plus the campaign source if the link you followed had one. The website keeps these two details in your browser for the current tab only (session storage) and sends them only when you submit a form. Cloudflare also tells us the country the request came from.

The request is emailed to our team through Resend and used to prepare your report or demo, to contact you about it, and to understand which pages bring requests. We keep these requests until you ask us to delete them (see Contact).

Booking a Demo Time

After sending the demo request form you can choose "Pick a time now". Only then does our scheduling page from Cal.com load on the page, with your name, email, and answers filled in. If you book a time, Cal.com processes the booking under its own privacy policy and adds the meeting to our calendar. Cal.com is not loaded unless you choose to pick a time.

Fonts and Translation

The website serves its fonts from our own servers; no request goes to Google Fonts. The website is translated with Loqalit Translate, as described in the Loqalit Translate section above.

Data Processing

Local Processing

  • Page scanning happens entirely in your browser
  • Scan history is stored locally in IndexedDB
  • Quality checks and MQM scoring run client-side

Server Processing

When you click "Analyze with AI":

  • Extracted text blocks and the page URL are sent to our Convex backend
  • Our backend builds prompts server-side and calls Google Gemini
  • Analysis results are returned to your browser, and for signed-in analyses the finding contents are stored in our backend database

Note: When R2 storage is configured, full request payloads (prompts and raw text blocks) may additionally be stored for quality review. Request metadata (page URL, domain, block count, token usage) is always logged.

Third-Party Services

ServicePurposeData receivedPrivacy Policy
ClerkAuthenticationEmail, sign-in session dataclerk.com/legal/privacy
ConvexBackend infrastructureAI analysis payloads, request logs, synced scan history, workspace/project context, invitations, and billing recordsconvex.dev/legal/privacy
Cloudflare PagesAuth UI and website hosting, website formsStandard web request data; website form submissions on their way to uscloudflare.com/privacypolicy
Google GeminiAI analysis, the platform assistant, and dictationExtracted page text and screenshots, text submitted through the AI-assistant connector, assistant questions and the project data needed to answer them, and dictation recordings (via our backend)ai.google.dev/terms
OpenRouterMachine translation (Loqalit Translate)Website text strings to be translated (via our service; no visitor identifiers)openrouter.ai/privacy
CrowdinOptional issue handoff (user-run)Finding text, page URL/title, page-text excerpt, screenshots — sent to your own projectcrowdin.com/page/privacy-policy
Phrase StringsOptional key-comment handoff (user-run)Finding text, page URL/title, page-text excerpt, screenshots — sent to your own projectphrase.com/legal/privacy
Dodo PaymentsPayment processing (merchant of record)Name, email, billing address, and payment details you enter on Dodo's hosted checkout; Loqalit never receives card datadodopayments.com/privacy
OneSignalTransactional email deliveryRecipient email address and the content of workspace/billing notification emailsonesignal.com/privacy_policy
HetznerServers that run platform scans and agentsThe pages being scanned and what the workers read from themhetzner.com/legal/privacy-policy
GitHub Actions and BlacksmithBuild runners that run platform scans and mobile app checksThe pages and app screens being checked and what the workers read from themGitHub privacy statement, blacksmith.sh/privacy
Cloudflare R2File storageScreenshots and stored analysis payloadscloudflare.com/privacypolicy
SlackOptional assistant in your Slack workspace (connected by your admin)Questions asked of the Loqalit app and its answers, which may include findingsslack.com/trust/privacy/privacy-policy
Google Analytics 4Website analytics, only if you allow it in the cookie bannerPages viewed, referring site, approximate location, device/browser, interactions; _ga cookiespolicies.google.com/privacy
ResendDelivery of website form requests to our teamThe contents of the free QA report and demo request forms, first page visited, referring siteresend.com/legal/privacy-policy
Cal.comDemo scheduling, only if you choose to pick a timeName, email, form answers, and the booking you makecal.com/privacy
AI assistant you connect (e.g. Claude by Anthropic, ChatGPT by OpenAI)Runs Loqalit tools you ask it to use (your choice, not our service provider)The results of the Loqalit tools it calls: project names, findings, scan status, linksYour assistant provider's policy, e.g. anthropic.com/legal/privacy, openai.com/policies/privacy-policy

We do not sell user data. The Extension does not include analytics or ad tracking libraries, and its production manifest allowlists no telemetry endpoints (Clerk SDK telemetry is disabled). Our website uses Google Analytics only if you allow it (see "Our Website").

Permissions

The Extension requests the following permissions:

PermissionPurpose
scripting, activeTabInject the content script to scan the current page
storageSave preferences and scan state locally
sidePanelDisplay results in Chrome's side panel
cookiesRequired by the Clerk authentication SDK for session synchronization (not used to read site cookies)
Loqalit backend/auth hostsHost permissions for our Convex backend and Clerk auth domains (AI analysis, sign-in)
api.crowdin.com (optional)Requested only when you set up the Crowdin integration
<all_urls> (optional)Requested only when you grant site access for scanning, or when you click "Allow screenshot access" in the export dialog to make screenshot evidence capture reliable. Revocable anytime in Chrome's extension settings

Data Security

  • All server communication uses HTTPS
  • Authentication tokens are not stored — a fresh short-lived JWT is fetched per request
  • We never store passwords (handled by Clerk)
  • No LLM API keys are stored client-side

Your Rights

You can:

  • Clear local data: Use the Reset button in the panel header or uninstall the extension (this does not delete server-side copies of synced or analyzed data)
  • Disable scan-history sync: Settings → Account → Scan history sync
  • Sign out: Dashboard Settings → Account section → Sign out
  • Disconnect an AI assistant: remove Loqalit in the assistant's settings, or revoke the API key it uses
  • Request deletion: Contact us to delete your account and associated server-side data

Data Retention

  • Local data: Retained until you clear it or uninstall the extension
  • AI request logs: Retained for 180 days, then deleted by a scheduled backend cleanup job
  • Product diagnostics: Detailed backend diagnostic events are retained for 30 days; aggregate rollups are retained for 180 days
  • Analysis findings, synced scan history, workspace/invitation records, and billing metadata: Retained while needed to provide the account and its history, or until deletion is requested where applicable
  • R2 payloads: Retained when configured, for quality improvement
  • Account data: Managed to support access and quota enforcement
  • Platform screenshots (web and mobile): Deleted after the workspace's retention period, 90 days unless agreed otherwise
  • Scan sign-in details and integration tokens: Kept encrypted until you remove them in the app or ask us to delete them
  • Website analytics (Google Analytics, only with consent): 14 months
  • Website form requests (free QA report, demo): Until you ask us to delete them

No Tracking or Ads

In the Extension, Loqalit Translate, and the AI-assistant connector, we do not:

  • Use analytics or ad tracking scripts
  • Display advertisements
  • Sell or share data with third parties beyond service providers listed above

Our website uses Google Analytics only after you allow it in the cookie banner, shows no advertisements, and has advertising features of Google Analytics turned off in its consent settings.

Children's Privacy

This extension is not intended for users under 13 years of age.

Changes to This Policy

We may update this privacy policy. Changes will be reflected in the "Last Updated" date above.

Contact

For privacy questions or data deletion requests:

  • Email: info@loqalit.com
  • Website: https://loqalit.com

Questions about this policy?