Everything we get asked

Click “Add to browser” to install from the Chrome Web Store, then pin the Deepread icon in your toolbar.

For the development build, run npm run dev in the repo and load .output/chrome-mv3 as an unpacked extension. Manually loaded builds do not auto-update.

Any Chromium browser: Chrome, Edge, Brave and Arc are the ones we test. The extension is Manifest V3, so a browser that has dropped Manifest V2 support still works.

Not for local tab classification — that runs in the extension without signing in. You need an account to keep a library across devices and to read saved pages on your phone.

Yes, at the moment. There is no paid tier. If that changes, it will be announced on this site before anything is put behind a payment.

Where the article text can be extracted, it is saved. Where it cannot, only the page info — title, address — is kept, and the row is marked “Page info only” in your library.

The distinction is deliberate: better to show you which items will not read later than to pretend everything was captured.

The article text could not be extracted from that page — often because the site renders its content in a way the extractor cannot follow, or blocks it. The title and address are still saved so the row stays findable.

Yes. Suggested categories are a starting point, and you can correct one from the library. Your correction is what gets stored.

Deepread looks at what you saved in a 24- or 48-hour window and groups pages that keep returning to the same idea across different sources. Each topic carries a grade — noise, weak or strong — based partly on how many independent sources back it.

Search runs over loaded pages: titles, summaries and categories. It cannot yet match words inside the article body.

Full-text search and pagination are scheduled for Wave 04 — the body index and the API’s total/cursor fields have to land first.

The interface ships in six languages: English, Chinese, Japanese, Korean, French and German.

Pinyin search over Chinese titles is available — you can type Latin letters only and still find the page.

Reading progress is set by hand on each page; your account remembers the value.

It does not record your position as you scroll. Automatic progress needs a different storage and sync design, and is not in this batch.

No. It is switched off while its storage cost is reworked — the previous design wrote a full snapshot on every change, which was too expensive to run. Existing data is kept, and the feature will come back once writes are incremental and quota-enforced.

No. The extension reads the current page when you ask it to save, and reads the tabs you ask it to organize. It does not collect your browsing history, does not inject ads and does not track you across sites.

In Deepread’s Cloudflare account, isolated per user. There is no end-to-end encryption — that is stated plainly on the privacy page rather than hinted around.

Not yet. Export is scheduled for Wave 04, in both JSON and Markdown.

Until then the answer stays “no”, not “coming soon”.

Not through the interface yet. Email the address on the privacy page and it will be handled by hand. A self-service delete control is planned.

The repository is public on GitHub. Issues and pull requests are read; there is no formal contribution process yet.