Explore Praxis
Download Praxis

A more capable browser. From Xasri.

Praxis Privacy Policy

Last updated: 18 November 2026

This policy explains how Praxis handles information when you use the browser, XASRI-powered features, and privacy or support services on praxis.xasri.com. It also explains your choices and the differences between information on your device, XASRI account records, visited websites, and external providers.

On this page

1. Who operates Praxis

Praxis is operated by its founders under the operating name Xasri Industries. This operating name does not represent Praxis as an incorporated company. The founders are responsible for the product's privacy practices.

For privacy questions, account-data requests or complaints, email Contact@xasri.com. This is our privacy and support contact; it is not presented as an appointed statutory data protection officer or regional representative.

2. Information on your device

Praxis stores regular tabs, browsing history, bookmarks, reading-list items, recent searches, shortcuts, settings, download records and saved AI conversations in your local browser profile. These records provide browsing, restoration, saved-content and conversation features.

History recording and saved AI conversations are enabled by default. You can change these settings and clear supported categories in Privacy Center. History is capped at 500 entries, recent searches at 12 and download records at 200. Other ordinary saved records remain until you remove them; there is no general automatic expiry for every local record.

Requested checks, reminders, workflows, research collections, snippets and tab handoffs are also saved locally. Scheduled checks may revisit a chosen website while Praxis is open and active. Those visits send requests to that website. They do not provide monitoring while the browser is closed.

Saved form profiles, form drafts and password-vault entries use platform secure storage. Password-vault values and saved form values are not offered to AI tools. Browser history and bookmark synchronization are not implemented: signing in to XASRI does not synchronize your local browser profile.

Ordinary profile records may be accessible to someone using the same operating-system/browser profile. Downloaded files are separate from download-history records.

3. Websites, cookies and permissions

When you visit a website, it receives ordinary network information such as your IP address, browser/runtime information, requested URLs, searches and information you submit to it. Website cookies, caches and site storage are handled by the browser runtime. The visited website's own privacy practices apply to that information.

Supported website permissions ask before access and require an identified requesting origin and an active page. Recorded grants apply to the current session. Privacy Center lets you revoke supported grants and reload affected pages. Unsupported or unverifiable requests are denied. Unsolicited pop-ups are blocked and certificate checks remain enabled.

Clearing cookies can sign you out of visited websites. Clearing download history does not delete files from your device. Available website-storage clearing depends on the platform and is described in the clearing interface.

The Praxis website's hosting and delivery services also receive ordinary request metadata when you visit the site. If you contact us, we use the email address, message and other information you provide to respond and handle your request.

4. Your choice to use XASRI

Cloud assistant and native XASRI Search requests are sent through XASRI Cloud at api.xasri.com. Praxis shows the current processing disclosure during Getting Started, with Allow and Not now, or before first use when current approval is missing.

Choosing Allow records the approved notice version and decision time locally. Choosing Not now during Getting Started disables cloud AI without granting approval. Skipping setup, signing in or completing the general introduction does not grant this approval. After approving the current notice, ordinary requests do not repeat that same notice every time. A changed notice or missing saved approval can require another choice.

The native XASRI Search page and its search transport check the same cloud preference and current approval before sending a query. Turning cloud AI off blocks these requests. This does not prevent ordinary website browsing or searches through another selected search engine.

Page text and URLs, selected text, selected-tab context and uploaded documents have separate sharing controls. These controls are off by default. Relevant page/text permissions are requested when you choose the associated action; document reading asks before each selected file upload. Sharing one category does not automatically allow the others. Disabling page sharing still permits ordinary browsing and approved prompt-only chat.

5. AI information and personalization

An approved request can include your prompt, relevant conversation context and the page, text, tab or file content you explicitly choose to share. XASRI Cloud processes this information to answer questions and provide requested research, coding, document-reading, speech or browser-assistance features.

Sarvam AI is used for supported cloud AI features. Search and research may use Tavily, Exa, Brave Search, public reference sources or a configured crawling service. Queries or requested URLs may go to the services needed for that feature. Not every provider receives every request.

For signed-in requests, XASRI can derive and save account-linked facts, preferences or goals to personalize later answers. These account memories are separate from conversations remembered locally on your device. Turning off local remembered chats does not, by itself, delete cloud account memories.

We do not use your personal data, submitted content or derived account memories to train our own AI models. Personalization using permitted facts or preferences is separate from model training. External providers have their own terms and processing arrangements, described below.

Webpages, search results and files cannot grant permissions or confirm browser actions. Supported agent actions use permission and review controls. Sensitive passwords, one-time codes, CAPTCHA responses, payment codes and identity-verification fields remain manual. Source filtering and redaction cannot reliably identify every secret that a user pastes into free text or a document.

6. External providers, training and retention

Provider handling depends on the applicable service, customer agreement and account configuration. A provider's general public policy is not proof that a particular account option is enabled.

Sarvam's published policy permits content use for improvement or model training unless an applicable opt-out is used. It describes a configurable content-retention default of 30 days after last access; account and technical records have different periods. Tavily's published terms permit query improvement unless the customer contract specifies otherwise, and its retention depends on service needs and applicable requests or exceptions. A no-training opt-out and exact retention settings for the Sarvam and Tavily accounts used by XASRI have not been confirmed. We therefore do not promise that requests to these services are excluded from provider improvement/training or deleted immediately.

Where an Exa request is covered by our zero-data-retention arrangement, that arrangement governs Exa's handling of that request. Its coverage and exceptions are specific to the applicable service and agreement. It does not cover Sarvam, Tavily, XASRI's own records or infrastructure logs/backups. Zero data retention is also distinct from a separate no-training commitment. We do not make a blanket no-training claim for every Exa product or request route.

Provider-held copies, processing jobs, logs and backups have their own retention and deletion boundaries. Clearing local data or receiving an AI response does not establish deletion from those systems. External processing is not processing entirely on your device.

7. Documents, images and voice

When you approve document reading, a selected PDF or supported PNG, JPG, JPEG or WEBP image and its filename can be sent through XASRI to Sarvam for processing. Extracted text may be used in a later requested AI answer. The document-reading flow does not establish immediate deletion of every provider job or copy after a response.

Voice input uses your device's platform speech-recognition service. That service may send microphone audio for external recognition; Praxis does not require recognition to happen entirely on the device. Recognized text goes to XASRI when used for a requested cloud feature. Microphone use requires the operating system's permission.

Requested Sarvam speech synthesis sends answer text, language and voice settings through XASRI and returns generated audio. This is different from uploading a microphone recording. The app removes its temporary playback files after playback, stop or failure. Device speech playback uses the installed voice engine, whose network behavior depends on that engine.

8. Search, Discover, icons and translation

Ordinary searches go to your chosen search engine. Native XASRI Search follows the approval controls in section 4.

Optional typed-query suggestions send typed text to Google and may fall back to DuckDuckGo. Optional remote website icons send a domain to the website or Google's favicon service. Suggestions and remote icons are off by default and have independent controls. Cached icons may remain locally until the cache is cleared.

The regular homepage's Discover feed loads BBC India headlines and thumbnails from BBC hosts by default. Those hosts receive ordinary network metadata. These requests do not include your browsing history, typed searches, account tokens or website cookies. You can disable Discover from its menu. Incognito and Family browsing do not load it. Selecting an article opens its original publisher link.

Requested page translation first explains that Google Translate will receive the selected page URL and may fetch its public content. After approval, Praxis opens the translation URL; it does not send that website's cookies or form values in the translation request.

9. Accounts, service access and diagnostics

XASRI sign-in processes your email address, supplied name and account/user identifier for authentication, account management and service access. Supabase supports authentication and applicable account database/storage features. OAuth access and refresh credentials use platform secure storage.

Supported account services can store profiles, saved conversations, messages, research/coding task records, attachments and account memories. Not every browser request creates a saved cloud conversation. These account records currently have no general automatic expiry and are handled through the applicable record or account-deletion operation.

Account and guest-use counters support service access and quotas. A random guest device label can accompany cloud, report and allowance requests, including an allowance refresh when the account page is opened. Guest lifetime-budget records do not currently have automatic expiry or a self-service server-deletion control.

Service audits may contain an account identifier, request identifier, time, method, route, status and duration for operation and security. Hosting, database and platform services may keep separate logs or backups. Their retention is not controlled by local browser clearing and a single verified production retention period has not been established for all of them.

The browser does not implement payment or subscription checkout. It may display an account plan. Transactions on other websites are handled under those services' terms.

10. AI reports and safety records

Report AI content opens an in-app form. It sends the selected reason, optional details and only the answer excerpt you choose and can edit. You can submit a reason-only report. Signed-in reports use the account session; guest reports use a random budget label. The app confirms receipt only after a valid service receipt. Receipt does not promise immediate review or removal.

Application-managed report content and moderation notes expire after 90 days; hash-only report abuse receipts expire after 24 hours. Pruning depends on service operation or restart. Account-linked reports are included in account deletion. Other infrastructure copies have separate handling.

Local agent safety records contain action/tool categories, permission decisions, confirmation status, outcomes and timestamps. They exclude prompts, page content, arguments, URLs and form values. Their default retention is 7 days, with 0, 7 or 30 days selectable, a 500-event cap and pruning when the records are accessed.

11. Optional measurement and separately configured services

First-party search measurement is disabled in the current release configuration. Enabling it requires a configured service and your opt-in. When enabled, it records activity dates, browser search-dispatch counts, the selected engine, platform, app version, your optional broad region choice and a random installation identifier separate from your account.

Measurement does not include search text, page URLs, browsing history, private-tab or Family browsing events. The receiving server still sees ordinary network metadata. Failed-send events can wait locally for up to 7 days and 500 events. Turning sharing off clears pending events and stops collection. Delete shared measurement data requests deletion from the configured service and reports whether it was confirmed; backups and previously exported reports have separate handling.

A separately configured backend may use OpenAI, Anthropic, Google Gemini or Groq. These optional integrations are not a statement that they are enabled for ordinary requests in the current XASRI release. Optional workspace services use authenticated, owner-scoped access. Their process-local safety records are capped at 500 events and 24 hours; workspace containers have a two-hour lifetime, with earlier cleanup attempted when closed. Deployment logs and backups remain separate.

12. Private browsing

Private tabs and their conversations are excluded from ordinary history, restoration and saved chats. Bookmarks and files you explicitly save may persist. Cloud AI in private tabs is off by default and requires a separate per-tab decision when enabled, in addition to applicable cloud approval.

Windows uses WebView2 InPrivate storage. Mobile private-cookie isolation depends on platform capabilities; we do not promise independent mobile cookie storage for every private-tab configuration. Private browsing does not hide requests from websites, internet providers or network administrators. An approved private cloud request is still sent to an external service.

13. Children and Family browsing

On Android, Praxis asks for a broad age-group choice before the regular browser and its cloud services start. It stores that choice locally, not a birthday. Under 18 and Prefer not to say enter protected Family browsing; 18 or older allows the regular browser. Parents can also open Family browsing from Settings on supported devices.

Family browsing uses separate website storage and parent-approved HTTPS reading sites/resource hosts. It does not start XASRI accounts, guest AI identifiers, cloud AI/OCR, agents, voice, Discover, remote suggestions or remote icons. Scripts, forms, frames, pop-ups, uploads, microphone and location access are disabled. It does not record browsing history, and parents can clear its website data.

Approved websites still receive ordinary requests and their content can change. A parent-approved handoff of a download to another app occurs outside Family browsing controls. Protected actions use the Android device credential; device authentication is not legal verifiable parental consent. The local age choice is not verified age assurance or a certification that every website is suitable for a child.

14. Downloads and ad filtering

Downloads contact the chosen source and may use its website session for authorized files. Download records and downloaded files have separate deletion controls. File-type and risk checks do not constitute a configured malware/reputation scanner. Private files are not uploaded for scanning.

Android and Windows ad filtering applies Brave-engine matching and bundled EasyList/EasyPrivacy rules locally, with supported site exceptions. Filtering applies inside Praxis. It does not use a Praxis advertising SDK, paid allowlisting, replacement ads or affiliate-link injection. Visited websites may still display their own advertising.

Windows rules update with app releases. Android signed rule updates require a configured publishing endpoint and public key. Such update requests do not include browsing URLs, cookies, referrers, account tokens or installation identifiers; the update host can receive ordinary IP/network metadata. Family browsing uses local rules without a network rule-update request.

15. Retention and deletion

Different records have different retention periods. Clearing one category does not erase every device, account, website or provider record.

RecordsRetention and deletion scope
Local browser recordsHistory/recent-search/download-record caps described in section 2; other records generally remain until removed. Files and website storage have separate controls.
Cloud account profiles, saved content and memoriesNo general automatic expiry; applicable record/account deletion removes associated active application data.
Guest lifetime-budget recordsNo automatic expiry or current self-service server-deletion control. Contact us for applicable requests and handling.
Local agent safety records7-day default; selectable 0/7/30 days, capped at 500 events and pruned on access.
AI reports and moderation notes90-day application-managed expiry; hash-only report abuse receipts have a 24-hour expiry. Service operation/restart is needed for pruning.
Optional pending measurementUp to 7 days and 500 locally pending events; disabling sharing clears the queue. Server, backup and export handling are separate.
Search cachesThe shared search cache has a default five-minute expiry. Other retrieval freshness windows do not prove physical deletion at exactly that time.
Account/task deletion markersHash-only account references, deletion status/time and task markers are retained indefinitely to prevent deleted records being recreated by late work.
Provider jobs, infrastructure logs and backupsApplicable provider/service rules and configuration; local deletion does not establish immediate removal from these systems.

Signing out removes the app's account session and attempts token revocation. It does not delete the account, local browser history or accounts held with visited websites.

To delete an XASRI account, sign in recently, open Your account, then select Account deletion. Review the warning and confirm. The service reports success only after confirmed cleanup of the account and associated active application data across supported XASRI services. Account memories and associated owned records are included. Failed or incomplete requests are not reported as successful deletion.

For instructions and requests without the app, visit Account deletion, or email Contact@xasri.com from your account's sign-in email with the subject Praxis account deletion request. We may need to verify account ownership. Sending an email starts a request; it does not instantly delete the account. Do not send passwords or one-time codes.

Provider copies, backups, legal retention and previously exported copies have separate handling. Hash-only deletion markers do not contain the deleted conversations/files, but remain security records. Local history, bookmarks, secure-storage entries, cookies and downloaded files are separate device data. Service-process memory may be physically cleared on restart after access has been blocked. Universal immediate erasure is not promised.

16. Security and international processing

Praxis uses platform secure storage for credentials and sensitive saved values, authenticated account operations, scoped permissions, review of supported consequential actions, source filtering/redaction and the operating system/browser runtime's security protections. Cloud endpoints use HTTPS and certificate checks remain enabled.

These controls are not a promise of perfect security or that every possible provider, device, backup and storage route has the same encryption properties. We do not claim end-to-end encryption for AI requests.

Providers and hosting services may process information outside your country. Applicable locations, protections and retention depend on the services and arrangements used. We do not promise processing exclusively in India or exclusively on your device.

17. Your choices and privacy rights

Privacy Center lets you disable cloud AI and optional sharing, change local history/chat preferences, clear supported local categories, manage supported form profiles/drafts, revoke permissions, clear safety records and export supported regular browser records. Exports exclude credentials, cookie contents, private tabs and secure form values. Turning off remembered chats removes saved local conversations.

Depending on your location and applicable law, you may have rights to access, correction, deletion, withdrawal of consent, portability, objection, restriction or a complaint to the relevant authority. Request help at Contact@xasri.com. We may request information reasonably needed to verify ownership and identify the relevant records. The applicable law and processing scope determine the response requirements and any exceptions.

An optional country/region setting uses your broad choice, not precise device location, to show relevant information. Local privacy controls remain available regardless of that choice.

18. Provider privacy documentation

Recipients depend on the requested feature and enabled service. These links explain each provider's public practices; specific customer arrangements can differ.

ServiceRelevant usePrivacy documentation
Sarvam AISupported cloud AI, document reading and requested speech synthesisSarvam Privacy Policy
TavilySearch/research queries when routed to itTavily Privacy Policy
ExaSearch/research and requested content retrieval when routed to itExa Privacy Policy
SupabaseAuthentication and supported account database/storageSupabase Privacy Policy
Brave SearchSearch/research when configuredBrave Search Privacy Policy
GoogleOptional suggestions, favicons, translation and separately configured GeminiGoogle Privacy Policy
DuckDuckGoSelected search and optional suggestion/search fallbackDuckDuckGo Privacy Policy
BBCDiscover headlines and thumbnailsBBC Privacy Policy
WikimediaEncyclopedia/reference retrievalWikimedia Privacy Policy
MicrosoftWindows WebView2 and associated platform servicesMicrosoft Privacy Statement

Research may also retrieve requested information from public sources such as arXiv, Crossref, OpenAlex and NASA. A configured crawling service may receive requested URLs. Separately configured optional backend providers include OpenAI, Anthropic, Google and Groq. Their inclusion here does not mean ordinary Praxis requests are sent to all of them.

19. Policy changes and contact

We will update this page when our privacy practices change. The date at the top identifies the current text. Material changes to optional processing require updated information and a relevant choice before the changed transfer, where applicable.

Privacy and support: Contact@xasri.com
Privacy page: praxis.xasri.com/privacy

Praxis browser illustration: Familiar at first. More as you go. A new-tab page with search, AI and voice shortcuts.

It starts with your browser.

Praxis is your place to explore the web. Xasri AI brings an assistant into that browsing experience.

Interactive introduction.
Not a product recording.

Explore Xasri AI