Privacy Policy
Last updated: 2026-11-19
Shelfwise is a browser extension that organizes your bookmarks using a large language model (LLM) you choose, with your own API key. It has no backend server, no user accounts, and collects nothing.
What we collect
Nothing. The developers of Shelfwise do not receive, store or have access to any of your data. There is no telemetry, no analytics, no crash reporting and no advertising.
What is stored on your device
Everything the extension keeps lives in your browser's local extension storage (chrome.storage.local) on your computer:
| Data | Purpose |
|---|---|
| Settings, including your API key | Connecting to the provider you configured |
| Up to 5 bookmark snapshots | Undoing a reorganization |
| The last 20 smart-saves (title, URL, folder, confidence, time) | The "Recent saves" list with Undo / Move |
Your API key is never written to Chrome sync storage, never logged, and never sent anywhere except the provider you configured. Removing the extension deletes all of this data.
What is sent, and to whom
Data is sent only when you trigger an action (organize, smart-save, test connection), and only directly from your browser to the LLM provider you configured:
- Google Gemini (
generativelanguage.googleapis.com), Groq (api.groq.com), OpenRouter (openrouter.ai), DeepSeek (api.deepseek.com), OpenAI (api.openai.com) or Mistral (api.mistral.ai), whichever you selected, or - the https OpenAI-compatible endpoint you entered as a Custom provider.
What is included in those requests:
- Organizing: the title, URL and current folder path of each bookmark being organized, your existing folder names, and any optional instructions you typed.
- Smart-saving: the title and URL of the link being saved and your existing folder names. When you use the X/Twitter button, the post's text (up to 1,000 characters). When you save from a text selection, the selected text (up to 1,000 characters).
- Test connection: a tiny fixed test prompt.
What is never sent: your browsing history, open tabs, page contents, cookies, passwords, form data, or any identifier of you or your device beyond what your provider's API inherently receives (such as your IP address).
How your provider handles this data is governed by your agreement with that provider (for example Google's, Groq's or OpenAI's API terms and privacy policy; note that some free tiers may use your requests to improve their models). Please review them.
Permissions and why they are needed
| Permission | Why |
|---|---|
bookmarks | Read and reorganize your bookmarks; create smart-saved bookmarks |
storage, unlimitedStorage | Keep settings, undo snapshots and recent saves locally (snapshots of large libraries can exceed the default quota) |
contextMenus | "Smart-save" entries in the right-click menu |
notifications | "Saved to …" confirmations and error alerts |
activeTab | Read the current tab's title and URL when you click Smart-save |
offscreen, clipboardRead | Read a URL from the clipboard, only when you press the clipboard shortcut |
| Host access to the six preset provider APIs | Calling the LLM provider you selected |
| Optional host access (requested at runtime) | Only if you configure a custom OpenAI-compatible base URL |
Content script on x.com / twitter.com | Showing the Smart-save button on posts; it reads a post only when you click its button |
Children
Shelfwise is not directed at children and collects no personal information from anyone.
Changes
If this policy changes, the new version will be published in this file with an updated date. Because the extension collects no data, changes cannot affect data we hold about you; we hold none.
Contact
Questions or concerns: please open an issue at https://github.com/geentjegeni/shelfwise/issues.
Provider-specific notes
- Google Gemini (free tier): Google may use prompts (your bookmark titles/URLs) to improve its products, and human reviewers may read them. The free tier isn't available for use in the EU/EEA, UK or Switzerland; use a paid key there.
- OpenRouter (free models): OpenRouter doesn't log prompts by default, but the providers serving free models may log and train on them.
- DeepSeek: data is processed and stored in China; prompts may be used for training unless you opt out in your DeepSeek account.
- Mistral: you can opt out of training under Admin › Privacy.
- OpenAI: API data isn't used for training by default.
- "Refresh models" sends only a model-list request (with your key, except for OpenRouter's public list), no bookmark data.