Palimpsest browser extension: privacy
Last updated 26 September 2026. This page covers the Palimpsest browser extension. The Palimpsest Privacy Notice covers your Palimpsest account and the rest of the service.
In short. The extension helps you learn Russian on the web: it colours Russian words by how well you know them, opens Palimpsest's word popup and flashcards, and shows a captions panel on YouTube. To decide which words a page contains, it reads the page on your computer. What it sends to Palimpsest is words (mostly their dictionary forms), not pages: never the address of a page you read or its text, unless you press Save to Palimpsest on it. We do not sell your data, share it for advertising, or use it for anything except your learning in Palimpsest.
What happens on your computer only
- Language check. On each page the extension checks, in your browser, whether the page is in Russian. For any other page, it stops there, and nothing about it is sent or saved.
- Dictionary forms. On a Russian page, it works out each word's dictionary form with a dictionary that ships inside the extension (about 98% of words on a typical page).
- How you read. It notes, in your browser, how long each word was on your screen while you were active, how often it came into view, the reading pace of its paragraph, and how close it was to a word you looked up. Only these per-word summaries are ever sent (see below), never the timings themselves, scroll positions or clicks.
- Stored in your browser: your extension token, a copy of your word scores and due flashcards (refreshed every 6 hours), your extension settings, reading events waiting to be sent, and a list of the Russian pages you visited recently (site name and word counts only, shown in the extension's menu). Pages that were not Russian are not kept.
What is sent to Palimpsest, and why
Everything below goes only to Palimpsest's own server (palimpsest.tech), encrypted in transit.
| When | What is sent | Why, and what is kept |
|---|---|---|
| A Russian page has words the built-in dictionary does not know (about 1–2 in 100) | Those word forms, shuffled so they are not in page order, without your token. Words that look like names are never sent. | To find their dictionary forms. Not stored. |
| You connect your Palimpsest account (by pasting a token you create in Settings) | Your token, with the requests below. | So the extension acts for your account. Disconnecting in the extension revokes the token; you can also revoke it in Settings. |
| Every 6 hours, while connected | A request for your word scores, due flashcards and relevant settings. | They are downloaded to your browser. |
| Colouring a Russian page, while connected | The word forms on the page that your downloaded scores do not cover, shuffled and in batches. | So the server can score them exactly as the app does. Not stored. |
| You click a word, look it up, change its score, reject it or flag it for a deck | The word's dictionary form, and your action. | Recorded in your account as the app records it. To write a definition, the word (not the page) is sent to our AI provider, Anthropic. |
| You review a flashcard that popped up on a page | The card and your answer. | Saved as a review in your account. |
| You read without looking words up (on by default; can be switched off in Palimpsest Settings) | The dictionary forms of words you had on screen, and did not look up, with how often they appeared, the day, and the per-word summaries described above. On YouTube: the words of caption lines that played. | They count towards your word scores, as reading in the app does. They are sent in batches mixed across pages, so no batch is one page, and they never include the page, its address or the time of day. |
| You press Save to Palimpsest | That one page's address, title and text, or, for a YouTube video, its video id, title, channel and captions. | Saved to your Palimpsest library, so you can read or watch it in the app. Only on your press, only that page. |
What is never sent
- Your browsing history, and the address or text of any page you did not save.
- Anything from pages that are not in Russian.
- Keystrokes, form contents, passwords, screenshots, audio or your location.
YouTube
On a YouTube video with Russian captions, the extension reads the captions file that YouTube's own player downloads, and shows it in a panel beside the video. The captions stay in your browser. The words heard are counted as described above, and the video id and captions leave your browser only if you save the video.
Your choices
- Pause (in the extension's menu) stops everything: no reading, colouring, lookups or sending.
- Stop counting your reading but keep colours and lookups: Palimpsest → Settings → Browser extension.
- Disconnect revokes the token and deletes everything the extension downloaded from your browser. Removing the extension deletes all of its local data.
- To see or delete what is stored in your account, use the contact details and requests in the Palimpsest Privacy Notice.
Permissions the extension asks for
- Read pages you visit (content scripts on all sites): to check the page's language and colour Russian words, on your computer. Nothing is read from a page that is not in Russian beyond that check.
- YouTube: to show the captions panel and read the captions YouTube's player downloads.
- palimpsest.tech: to talk to your Palimpsest account.
- Storage, alarms: to keep your scores, settings and queued events, and refresh them every 6 hours.
- Context menu: adds "Look up this word in Palimpsest" to the right-click menu.
- Active tab: to show, in the extension's menu, what it did on the page in front of you.
Limited use
The use of information received through the extension adheres to the Chrome Web Store User Data Policy, including the Limited Use requirements. Data is used only to provide and improve Palimpsest's learning features for you. It is not sold, not transferred to third parties except the service providers Palimpsest needs to run (hosting, database, and the AI provider that writes definitions), not used for advertising, and not used to determine creditworthiness or for lending.
Third-party material
The built-in Russian dictionary is derived from OpenCorpora
(CC BY-SA 3.0) and the Russian word list of
wordfreq (CC BY-SA 4.0). Those dictionary files are
shared under the same licences, and the extension package lists them in vendor/DATA_LICENSE.md. The analyser
that reads them is Az.js (MIT licence).
Changes and contact
If this changes what the extension sends, we will update this page and say so in the extension. Questions: see how to contact us.