100% offline · Chrome extension

A private, searchable memory of everything you’ve seen in Chrome.

Q: Is it really 100% offline?

Yes, and it is enforced rather than promised. The build scans the service worker and the page content script for network APIs — fetch, XMLHttpRequest, WebSocket, EventSource and sendBeacon — and fails if any of them appears where it should not. No web fonts are loaded, no icons are fetched from a CDN, and the PDF reader is bundled in the package rather than downloaded at run time.

Q: Will it slow down my browsing?

Indexing runs in the background, in the extension's own worker, not on the page's main thread. The search index is designed for a sub-50 ms response across 100,000+ records, and lists use virtual scrolling so a large memory does not cost you a slow interface.

Q: Are private or incognito windows recorded?

No — not unless you explicitly allow the extension to run in incognito, which is a switch only you can turn on from chrome://extensions. Chrome does not let an extension grant itself that access. Pages Chrome protects, such as chrome:// pages and the Chrome Web Store, cannot be read at all.

Q: Can I get my data out again?

Export everything to JSON, CSV or Markdown at any time from Settings, and import it back on this machine or another one. You can pick which categories to include, delete any single record, or reset the whole database in one action. There is no lock-in because there is nowhere to be locked into.

Q: Does it capture passwords or card numbers?

No. Clipboard capture refuses password, one-time-code, CVV and card-number fields outright, so text copied from them is never stored. You can also add domains to a never-record blocklist, with wildcards, so a whole site or path is skipped entirely.

Q: How much disk space does it use?

It depends on how much you browse and what you capture. Page text is the bulk of it; a few thousand indexed pages is typically tens of megabytes. Settings shows the live figure, you can set a retention period so older records are removed automatically, and export supports omitting page text for a much smaller file.

Q: Is it free? Are there ads or upsells?

It is free, with no account, no card, no trial, no ads and no upsell. There is no server to pay for because there is no server.

Q: Does it work in Edge, Brave or other Chromium browsers?

It is built and tested against Google Chrome 116 and later. Most Chromium browsers — Edge, Brave, Vivaldi, Opera — can install extensions from the Chrome Web Store and should run it, but they are not part of the tested matrix, and browsers differ in their support for the side panel and offscreen APIs.

Q: What happens to my data if I uninstall?

Uninstalling removes the extension's local database with it, permanently. Nothing is retained anywhere else, because nothing was ever sent anywhere else. Export first if you want to keep your memory.

Browser history is a list of links you barely recognise. Digital Memory records what you actually read — pages, PDFs, downloads, notes — and makes it findable in milliseconds. It stays on your computer, because there is no server to send it to.

One search across page text, PDFs, bookmarks, downloads and notes
Your day as a timeline, grouped by hour, site or category
No account, no sync, no telemetry — enforced by the build, not promised

Add to Chrome — it’s free See it in action

Chrome 116 or later · No sign-up, no card, no trial · Uninstall takes your data with it

Digital Memory dashboard showing counts for pages indexed, visits recorded, bookmarks, downloads, notes, clipboard entries and indexed PDFs, alongside recent activity, most-visited sites, a file-type breakdown and a privacy panel reading 100% offline.

0

network requests made
0

accounts, sign-ups or cards
100%

stored on your own device
<50ms

search response target

The problem

You remember the sentence. Not the URL.

You read something useful three weeks ago. You remember roughly what it said, roughly when, and nothing else. Browser history cannot help: it indexes addresses and titles, keeps no text, and forgets in ninety days. So you search the web again and hope to find the same page twice.

How the usual options compare with a local index of your own browsing.
	Browser history	Cloud read-later & notes	Digital Memory
Searches the text of pages	No	Only what you saved	Yes, automatically
Searches inside PDFs	No	Sometimes, on upload	Yes, locally
Needs you to save things first	No	Yes	No
Data leaves your device	Synced to your account	Yes, to their servers	Never
Account required	Your Google account	Yes	None
Works with no internet connection	Partly	No	Always
Export in an open format	Limited	Varies	JSON, CSV, Markdown

Features

A second brain that fills itself in.

Twelve things it does, all of them on your machine. Every capture category can be switched off individually, or all at once.

One search box for everything

Pages, PDFs, bookmarks, downloads, notes and clipboard history in a single query. Fuzzy and partial matching handles typos and half-remembered words, and matched text is highlighted in the result.

Your day, by the hour

A timeline of today, yesterday, this week or this month, grouped by hour, website or category, with how long you actually spent. Find the afternoon, not the address.

Insights that are actually about you

Time on page, pages read, estimated reading time, activity by hour, top sites ranked by time, and where your attention went by category — daily, weekly, monthly or all time.

Notes that stay with the page

Write Markdown notes attached to the page you were reading, with live preview. Pin them, colour them, and search them alongside everything else.

Bookmarks you can finally find

Your existing bookmarks, mirrored and made searchable, with tags, favourites, sorting and filters. They are read, never modified.

Downloads with context

Filename, source address, file type, size and date — searchable, filtered by category, with a jump straight to the file in your file manager.

Clipboard history

Text you copy, kept with the page you copied it from, pinnable and searchable. Password, one-time-code, CVV and card-number fields are refused outright — that text is never stored.

Full-text search inside PDFs

Title, author, metadata and the text of every page, extracted locally with a bundled copy of pdf.js. Off by default — it is the one feature that reads a file, so you turn it on deliberately.

OCR, on your own CPU

Pull text out of images and screenshots so they turn up in search too. Tesseract runs inside the extension; no image is ever uploaded to an OCR service, because none is contacted.

Sorted into categories, by rules you own

Development, AI, finance, education, news, shopping, travel, health and more — classified by a rule-based system you can read and change. No model, no guessing, no API call.

Keyboard first

Quick search, open memory, new note and today’s timeline all have a shortcut, and / focuses search anywhere in the dashboard or side panel. The mouse is optional.

Your data, portable

Export to JSON, CSV or Markdown, choosing exactly which categories to include. Import it back on any machine, restore from a backup, or rebuild the index from your own records.

Screenshots

Dark by default, dense where it counts.

Four surfaces: the dashboard, search and timeline together, insights, and the settings that keep you in control. Screens show sample data.

The dashboard: eight counters across pages indexed, visits recorded, bookmarks, downloads, notes, clipboard, PDFs indexed and search terms; a recent-activity list; most-visited sites; a top-file-types doughnut chart; and a privacy card showing local data size, storage used and guarantees with tracking, cloud sync, analytics and telemetry all disabled. — **Dashboard** Everything your browser remembered, counted and cross-linked — with the privacy panel reporting local database size and a standing row of disabled: tracking, cloud sync, analytics, telemetry.

Search results for the query google across pages, bookmarks, downloads, notes, clipboard and PDFs with type filters and date filters, showing 116 results returned in 6 milliseconds; below it the timeline view of the same day grouped by hour with visit durations. — **Search and timeline** One query, every type, filtered by date — with the result count and the time it took stated plainly. Underneath, the same day laid out hour by hour.

The insights screen showing four hours thirty-four minutes on page, thirty-six pages, thirty-four minutes of reading, two items saved, an activity bar chart by hour of day peaking at midday, a doughnut chart splitting attention across categories, and the top five sites ranked by time spent. — **Insights** Where the time actually went: activity by hour, attention split by category, and the sites that took the most of it — today, this week, this month or all time.

The settings screen showing keyboard shortcuts, a never-record blocklist accepting wildcard domain rules, export controls for JSON, CSV and Markdown with per-category toggles, restore from a JSON backup, appearance density options, a reduce-motion switch, a minimum visit length control and a rebuild search index action. — **Settings** A never-record blocklist with wildcards, retention and minimum visit length, export and restore, density and reduced motion — and a rebuild button, because the index is derived and can always be regenerated.

How it works

Install it, then forget about it.

There is nothing to configure, connect or sign into. The useful part starts the moment you install and shows up the first time you cannot remember something.

Install

One click from the Chrome Web Store. No account, no onboarding, no permissions dialog beyond what Chrome shows you — and every one of those permissions is explained on this page.
Browse as usual

Pages you actually read are indexed in the background, on your device. Very short visits are discarded as pass-throughs, and anything on your blocklist is skipped entirely.
Recall in milliseconds

Press Ctrl Shift K, type the half-word you remember, and get the page back — with the sentence that made you keep it.

Privacy

Not a policy. An architecture.

Most privacy pages describe a setting. This one describes an absence: there is no server, no account system, no analytics and no telemetry anywhere in the extension — not switched off, but not written.

What we collect about you

Nothing.

That is the whole policy in one word. The rest is detail about what stays on your device, and how to take it with you or destroy it.

Enforced, not promised

The build scans the service worker and content script for fetch, XMLHttpRequest, WebSocket, EventSource and sendBeacon — and fails if any turns up where it should not.
No web fonts, no CDN icons; pdf.js ships inside the package.
The content security policy allows scripts only from the extension’s own package, so remote code could not run even if it were injected.
The privacy panel shows the live request count. It reads zero.

Never recorded

Passwords, one-time codes, CVVs and card numbers.
Incognito windows — unless you allow it yourself in chrome://extensions, which only you can do.
chrome:// pages and the Chrome Web Store, which extensions cannot read at all.
Any site on your blocklist, and anything you switch off.

One honest exception

Making a PDF searchable means reading the PDF. If you switch on PDF indexing, the extension reads PDFs you have already opened yourself and extracts their text locally — nothing is uploaded. It ships off, and indexing PDFs stored on your own disk additionally needs you to enable “Allow access to file URLs” on the extension’s card in chrome://extensions, which Chrome only lets you grant yourself.

Every permission, and why it exists

The twelve permissions in the manifest, plus host access. Nothing is requested that is not used.
Permission	What it is for
`bookmarks`	Read your bookmarks so they become searchable. They are never modified.
`downloads`	Read download details, reveal a file in your file manager, and save your export.
`storage`, `unlimitedStorage`	Hold your settings and your records. Unlimited, because Chrome’s default quota would silently start discarding your memory.
`tabs`, `webNavigation`	Know which page you are actually looking at, and for how long — the timeline is built from this.
`sidePanel`	Provide the side panel, so your memory sits beside the page instead of replacing it.
`contextMenus`	Right-click actions on your own selection — save a note, copy to memory.
`alarms`	Periodic local housekeeping: reconcile bookmarks and downloads, run data cleanup.
`offscreen`	Parse PDFs in a page context, which a service worker cannot do.
`clipboardWrite`	Put a saved snippet back on your clipboard when you click it.
`favicon`	Show each site’s icon from Chrome’s own local cache — not from the network.
`http`, `https`, `file` access	Read the text of pages you visit so you can search by what a page said. Reading happens inside the page; the result is stored locally and sent nowhere.

Read the full privacy policy

Keyboard

Four shortcuts and a slash.

All of them remappable from chrome://extensions/shortcuts. On macOS, Ctrl is ⌘.

CtrlShiftK Quick search
CtrlShiftM Open memory
CtrlShiftN New note
CtrlShiftT Today’s timeline
/ Focus search, in the dashboard or side panel

Specifications

The technical answer.

Version	0.1.0 · Manifest V3
Requires	Google Chrome 116 or later — Windows, macOS, Linux, ChromeOS
Price	Free. No account, no card, no trial, no ads, no upsell.
Storage	IndexedDB, in your own Chrome profile, via Dexie. Never synced.
Search	Local inverted index with fuzzy and partial matching. Built for a sub-50 ms response across 100,000+ records.
Bundled libraries	pdf.js for PDFs, Tesseract.js for OCR, Chart.js for insights — all shipped in the package.
Network	None. No remote code, no remote assets, verified at build time.
Export formats	JSON, CSV, Markdown — with a per-category picker and an option to omit page text.
Accessibility	Keyboard-first navigation, a reduce-motion setting, and two density modes.

Questions

Reasonable suspicions, answered.

Anything that reads your browsing deserves scepticism. Here is the detail behind the claims.

Is it really 100% offline?

Yes, and it is enforced rather than promised. The build scans the service worker and the page content script for network APIs — fetch, XMLHttpRequest, WebSocket, EventSource, sendBeacon — and fails if any of them appears where it should not. No web fonts are loaded, no icons are fetched from a CDN, and the PDF reader is bundled in the package rather than downloaded at run time.

Will it slow down my browsing?

Indexing runs in the background, in the extension’s own worker, not on the page’s main thread. The search index is designed for a sub-50 ms response across 100,000+ records, and lists use virtual scrolling so a large memory does not cost you a slow interface.

Are private or incognito windows recorded?

No — not unless you explicitly allow the extension to run in incognito, which is a switch only you can turn on from chrome://extensions. Chrome does not let an extension grant itself that access. Pages Chrome protects, such as chrome:// pages and the Chrome Web Store, cannot be read at all.

Can I get my data out again?

Export everything to JSON, CSV or Markdown at any time from Settings, and import it back on this machine or another one. You can pick which categories to include, delete any single record, or reset the whole database in one action. There is no lock-in because there is nowhere to be locked into.

Does it capture passwords or card numbers?

No. Clipboard capture refuses password, one-time-code, CVV and card-number fields outright, so text copied from them is never stored. You can also add domains to a never-record blocklist, with wildcards, so a whole site or path is skipped entirely.

How much disk space does it use?