You read a page last week and now you cannot find it. Backtrack indexes the pages you read and lets you search them by what they were about, so a query like “that article comparing electric cars by range” finds the right one even when you forgot its title.
What it does
Chrome history search matches titles and URLs. If you remember a page by what it covered rather than its exact title, that search comes up empty. Backtrack fills the gap. It reads the pages you visit, builds a meaning-based index on your device, and lets you search that index later by describing the topic.
Type what you remember, even loosely, and Backtrack ranks the pages whose content is closest in meaning. It handles synonyms and rough wording because it compares ideas rather than exact strings. A search for “the recipe with brown butter and sage” surfaces the page even if those words never appeared in its title.
All of this runs inside your browser. The page text it reads and the index it builds stay on your machine. The only network request Backtrack ever makes is a one-time download of the AI model. After that it works offline.
Features
Semantic search
Search by meaning, not exact words. Describe what a page was about and Backtrack ranks the closest matches, synonyms and typos included.
Fully on-device
The AI model runs in your browser via ONNX and WebAssembly. Page text and the search index never leave your machine.
Works offline
After a one-time model download from the Hugging Face CDN, Backtrack indexes and searches with no network connection at all.
Sensitive pages skipped
A built-in blocklist keeps email, banking, login pages, and local addresses out of the index. Form fields and passwords are ignored.
No account, no tracking
No sign-up, no servers, no analytics. Backtrack loads no remote scripts and sends none of your browsing anywhere.
Wipe anytime
Clear the entire local index from the search page whenever you want. Uninstalling removes everything Backtrack stored.
How it works
When you open a readable page, Backtrack extracts its text, turns it into meaning vectors with a small AI model, and stores them in your browser. A search runs your query through the same model and ranks the stored pages by closeness. Everything below happens inside Chrome.
all-MiniLM-L6-v2 model, which runs in an offscreen document via ONNX and WebAssembly so the browser stays responsive. The vectors and their source text are saved in IndexedDB on your device.
Read more: How to find a website you visited but forgot the name and how to search your browser history by page content.
Privacy
Backtrack is built so the sensitive parts of your browsing never get captured. It skips a built-in blocklist of surfaces and ignores anything you type into a page.
Nothing Backtrack reads is transmitted. The index lives in your browser, and you can wipe it from the search page at any time.
FAQ
Is my browsing uploaded anywhere?
No. The page text Backtrack reads and the index it builds stay on your device. The only network request it makes is a one-time download of the AI model from the Hugging Face CDN. That request fetches model files only, not your pages, history, or queries. Read the full privacy policy.
How is this different from Chrome history search?
Chrome history matches titles and URLs. Backtrack indexes the body text of pages and searches by meaning, so it finds a page from a description of its content even when you do not remember the title or any exact word from it.
Does it work offline?
Yes, after the first run. The AI model downloads once, then indexing and search run entirely on your device with no connection needed.
What about sensitive pages like email and banking?
They are skipped. Backtrack has a built-in blocklist covering email, banking, login and sign-in pages, and local addresses such as localhost. It also ignores form fields and passwords on every page.
Can I delete what it has stored?
Yes. You can wipe the entire local index at any time from the search page. Uninstalling the extension removes everything it stored in your browser.
Which AI model does it use?
It uses all-MiniLM-L6-v2, a small sentence-embedding model that runs in your browser through ONNX and WebAssembly. It is compact enough to index pages quickly without slowing the browser down.
On-device. No accounts. No tracking. Page text and index stay in your browser. Read the privacy policy
Free to install. Starts indexing as you browse, searchable in seconds.
Install for free ->