Last updated: 2026-06-30

This document has two parts: an English version and a Korean (한국어) version. They say the same thing. Host either or both at a public URL and paste that URL into the Chrome Web Store listing's "Privacy policy" field.

English

The short version

Recall keeps everything on your own device. It does not send your browsing, the pages you read, your searches, or anything else to us or to any other server. We do not have a server that receives your data. We collect nothing. We sell nothing. There are no accounts, no logins, no analytics, and no ads.

What Recall does

Recall helps you find pages you have already read. When you genuinely read a web page, Recall saves a clean copy of that page's readable text into a private database that lives inside your browser, on your computer. Later you can search those saved pages by meaning, in your own words, in Korean or English. The search runs on your device too.

To do this, Recall uses two things that both run locally:

  1. A small database (SQLite, stored in your browser's private OPFS storage).
  2. A bundled language model (~107 MB, IBM Granite multilingual embeddings) that turns text into searchable "meaning vectors." The model file ships inside the extension. It is never downloaded at runtime.

Where your data lives

All captured text, page titles, page URLs, and the search index live only in your browser's local storage on your machine. None of it is uploaded. Recall's network policy is enforced in code: the extension's Content Security Policy is connect-src 'self', which means the extension is technically blocked from connecting to any remote server.

What is captured — and what is NOT

Recall is careful about what it saves. Capture is gated: a page is only auto-saved after you have actually engaged with it (it has been visible for a dwell period and you scrolled through a meaningful part of it or selected some text — short pages that fit on screen count as read).

Recall does not auto-save: