Privacy policy for Raindrop AI Assist
Raindrop AI Assist by agunnell
Privacy policy for Raindrop AI Assist
Effective Date: October 19, 2026
Last Updated: October 19, 2026
Raindrop AI Assistant ("the Extension", "we", "our") is an open-source Firefox WebExtension designed to provide AI-powered collection routing, contextual auto-tagging, and smart summaries for Raindrop.io.
We believe privacy is a fundamental right. This extension is engineered under a strict Zero-Telemetry, Client-Direct Architecture: it operates entirely within your browser, communicates solely with the official services you configure, and contains no analytics, no trackers, and no intermediary servers.
- No Developer Servers: There is no backend, proxy, or relay server operated by the extension developer. All network requests travel directly from your browser to your chosen service endpoints.
- Zero Telemetry or Analytics: We do not collect, monitor, log, or sell any usage data, browsing history, telemetry, or personal identifiers.
- Local Storage Only: All credentials (API tokens, keys) and configuration preferences remain exclusively inside your browser's private local extension storage (
browser.storage.local). - On-Demand Activation Only: The extension only inspects the active tab when you explicitly activate it (by clicking the toolbar icon or pressing <kbd>Alt</kbd> + <kbd>R</kbd>). It never runs in the background across passive browsing sessions.
- Raindrop Personal Test Token: Required to authenticate directly with the official Raindrop.io API to query collections/tags and create bookmarks.
- AI Provider API Keys: Optional keys for Google Gemini or OpenAI (not needed if using a Local LLM).
- Storage Location: Stored exclusively in Firefox's sandboxed local storage (
browser.storage.local). These keys never leave your device except as bearer tokens in direct HTTPS authorization headers to the respective API endpoints.
When you trigger the extension on an active tab, the content script extracts:
- Page Title and Canonical URL
- Meta Description (from OpenGraph / Twitter / HTML meta tags)
- Cleaned text content from the article/main body (noise elements such as scripts, styles, navs, headers, footers, and ads are stripped)
- User-highlighted text (if any text is currently selected on the page)
[!NOTE]
The extracted text is truncated to ~6,000 characters to optimize token usage. This text is passed directly to the AI provider in-memory and is never logged to disk or stored permanently by the extension.
To optimize classification speed and minimize API requests, the extension caches your Raindrop collection names, hierarchy IDs, and existing tags in
browser.storage.local. You can view counts or refresh this cache at any time from the Extension Options page.The extension only communicates with external endpoints that you explicitly configure:
| Service / Endpoint | Data Sent | Purpose | Privacy Policy |
| :--- | :--- | :--- | :--- |
| Raindrop.io API<br>
https://api.raindrop.io/v1/* | Raindrop Test Token, URL, page title, summary notes, selected collection ID, and tags | Fetch collections/tags, check URL duplication, and save bookmark | Raindrop.io Privacy Policy || Google Gemini API<br>
https://generativelanguage.googleapis.com/* (Optional) | Gemini API key, collection names, existing tags, page title, URL, meta description, and article excerpt | Analyze page context and return structured JSON (collection ID, tags, summary, bullet points) | Google Privacy Policy || OpenAI API<br>
https://api.openai.com/v1/* (Optional) | OpenAI API key, collection names, existing tags, page title, URL, meta description, and article excerpt | Analyze page context and return structured JSON | OpenAI Privacy Policy || Local LLM Server<br>
http://localhost/*, http://127.0.0.1/* (Optional) | Collection names, existing tags, page title, URL, meta description, and article excerpt | Local inference via llama.cpp, Ollama, or custom servers. 100% local — no data leaves your computer. | N/A (Self-hosted on-device) |We do not use any third-party tracking scripts, CDNs, error reporters (e.g. Sentry), or analytics packages (e.g. Google Analytics).
In accordance with Mozilla Add-on Policies, here is why each permission in
manifest.json is strictly necessary:| Permission | Technical Requirement & Scope |
| :--- | :--- |
|
activeTab | Grants temporary access to the currently active tab when you click the extension action or trigger the keyboard shortcut, allowing the extension to inspect the URL and title. ||
scripting | Used to inject scripts/content.js on-demand into the active tab to extract article text and highlighted quotes for AI summarization. ||
storage | Provides access to browser.storage.local to securely persist your API tokens, AI model preferences, custom rules, and cached taxonomy on your local machine. ||
webRequest & webRequestBlocking | Restricted strictly to local IP patterns (http://localhost/* and http://127.0.0.1/*). It rewrites the browser Origin header to http://localhost and appends permissive CORS headers so that in-browser fetch requests to local inference daemons (such as llama-server or Ollama) succeed without 403 Forbidden or CORS errors. It never touches external web traffic. ||
host_permissions | Allows direct API calls from the popup and options pages to api.raindrop.io, generativelanguage.googleapis.com, api.openai.com, and local loopback addresses (localhost / 127.0.0.1). |- Modifying or Clearing Keys: You can update, overwrite, or delete your Raindrop Token and AI API Keys at any time in the Extension Options page.
- Complete Data Purge: Uninstalling or removing the extension from Firefox immediately and permanently purges all stored credentials, preferences, and taxonomy caches from
browser.storage.local. - Third-Party Retention: Data sent to Raindrop.io, Google, or OpenAI is governed by their respective retention and API data-usage agreements. (Note: Under standard Google AI Studio and OpenAI API platform policies, data submitted via paid/standard APIs is generally not used for model training; refer to their developer terms for details).
- COPPA & Children's Privacy: The extension does not knowingly collect or target information from children under 13.
- GDPR & CCPA: Because we operate no servers, collect no personal information, and track no user activity, we do not process, monetize, or transfer your personal data.
The code for Raindrop AI Assistant is open source. You can audit every network call, content extraction script, and storage interaction directly in the repository source code.
If you have questions about this Privacy Policy or wish to report a security/privacy concern, please open an issue on the project's GitHub repository or contact the maintainer directly through the project repository.