Your library, your choices.

Updated October 16, 2026. This describes the development build. Connected services operate only when the project owner configures them.

Capture

Boomerang saves links, titles, selected/readable text, notes, topics, progress, feedback and uploaded originals. When automatic article capture is enabled, saving a public link requests its readable content from the source website. Login cookies are not sent, and private network targets are rejected. The extension captures only through explicit toolbar or context actions. Optional bookmark and tab imports request permissions when used. It does not continuously collect browsing history or intentionally capture form entries.

Local mode

Metadata uses browser local storage or Chrome extension storage. Files use IndexedDB. Web and extension local libraries are separate. Local capture alone sends no content to AI services. When the optional local AI connection is enabled, readable source text is sent through your configured backend to OpenRouter for automatic summaries and category assignment, including while signed in. Signed-in results are saved to your account using your authenticated permissions. Website analytics are optional; see the analytics section below. Local storage is not an encrypted vault; other users of the browser profile may access it.

Local image and scan OCR runs on your development server using Tesseract, which downloads English language data on first use. For uploaded images without readable text, the app may send the image through your local backend to OpenRouter’s free vision routing. Visual interpretations can be inaccurate and are labelled. PDF scan pages use OCR only. Extracted text remains saved for search but is hidden in the reader. Rabbit-hole topic links share the selected search query with Google only when opened.

Connected accounts

Signed-in account metadata and files use the configured Supabase database and private Storage bucket. Account policies restrict access. Original files use short-lived signed links. Account saves synchronize across signed-in web and extension sessions. Existing local saves are uploaded only through an explicit import.

AI and search

In connected mode, new source text may be queued for OpenRouter processing. Text goes to OpenRouter and its model provider; OCR sends the selected uploaded image. AI memory search can send titles, concepts and source excerpts. External discovery sends title and concepts to the configured search engine. Provider retention policies apply. Do not capture sensitive material unless this processing is appropriate. Live search requires operator configuration and may incur charges.

Controls

Use local mode, sign out, edit notes/topics, archive, dismiss or snooze suggestions, and export metadata. Backups include text but not file bytes; keep originals separately. Archiving is not deletion. Complete cloud export and account deletion are not yet available in the app; the operator must provide them before launch. Clearing browser data or uninstalling removes local copies.

Before public distribution

The publisher must provide support, deletion and retention policies, provider-specific disclosures, and verified account isolation. This development disclosure is not a production privacy certification.

Back to Boomerang

Automatic text-selection button: Boomerang runs a bundled content script on regular HTTP/HTTPS pages where Chrome grants site access. It observes selections locally to position a save button. Selected text is saved only when you click the button. It avoids editable text fields and initializes automatically when regular web pages load. Saves remain local unless you are signed into a connected account.

Copy earlier saves uploads saved content and available local original files to the signed-in Supabase account only when you choose that action. Local copies remain on the device.

Email and password authentication is handled by Supabase. New saves in the configured app require an account and are stored in that account; previously saved local data is retained for explicit recovery.

AI fallback and topic organisation

When configured, unsuccessful OpenRouter summary requests are retried through River. Captured source text may be sent to River and its model provider. Background topic checks may send titles, summaries, concepts and up to 3,000 characters of source text per unresolved item to these providers. They reuse existing categories or suggest a new heading; manual topic choices are preserved.

Optional website analytics

On boomerangit.me, Microsoft Clarity loads only after you allow analytics. It processes interaction and technical usage data for heatmaps and session recordings, with page content explicitly masked. Advertising storage is denied. We do not supply your account identity or saved content as custom analytics events. Choose Analytics settings on the website to change your decision. Clarity is excluded from the Chrome extension and local development. Microsoft processes analytics data under its own privacy terms.

Support and privacy requests

Contact boomerangitsupport@gmail.com for product help, account questions, or to request access to or deletion of your account data. Send requests from the email associated with your Boomerang account. Do not send passwords or API keys. Requests are handled manually; archiving a save does not delete it.

Optional payment testing

Beta access is free. Approved testers may explicitly open a Dodo Payments test checkout. This sends their account email and account identifier to Dodo to associate the test session. Card or billing details entered on Dodo are handled by Dodo; use only test payment details. Test checkout does not activate a paid Boomerang plan. No saved articles or passages are sent for checkout.