Memory & search

Trace indexes every page you visit without you doing anything. No manual bookmarking, no "save for later" button to remember to click.

What gets indexed

For each page you visit, Trace stores:

• The full extracted text of the page.
• A semantic vector embedding, used for meaning-based search.
• An AI-generated summary.

This happens automatically as you browse. The Memory Indexer background agent keeps the index up to date without any manual action.

Searching your history

Open the chat sidebar (Ctrl/Cmd+K) and ask in plain language, for example:

show me that article about vector databases I read last month

Trace's memory search understands relative time expressions like "yesterday," "last week," "last 3 days," and "last month," and ranks results by semantic relevance, not just keyword matches. This means you can find a page even if you don't remember the exact words it used.

There's also a dedicated full-text and vector hybrid search across all indexed pages for more precise lookups.

Tip:

If you can describe what a page was *about* rather than its exact title, memory search will usually find it. That's the point: it searches meaning, not just text.

Browsing History

For a simpler, point-and-click view instead of a search query, open History from the nav sidebar. It shows your full history grouped by day, with visit duration, AI summaries, and one-click navigation back to any page. See Browsing History for details.

Next steps

• Personal memory covers the separate system that extracts facts *about you*, not just page content.
• Knowledge Graph shows how entities and relationships extracted from pages connect to each other.