Full-Text Search Across Everything You've Saved (vs Browser Bookmarks)
Cowpin
10/18/2026
Chrome and Safari search title and URL. That fails the moment you remember a phrase inside the article, not the headline you glanced at when you saved.
Cowpin is built for “find the paragraph” — but only when the page text actually lives in your library.
What gets indexed
- Title and URL — always.
- Tags and notes — your vocabulary and paste-ins.
- Archived page text — when Cowpin fetched and stored a readable snapshot (
searchVector/ FTS in Postgres). - Semantic search — embeddings over saved content for “meaning-close” recall (product tier / fair-use limits apply).
Browser bars never see (3). Read-later apps that only store URLs without archive behave like fancy bookmark lists until something re-fetches the page.
How to get searchable saves
| Path | Search quality |
|---|---|
| Extension / quick-add on live pages | Best — archive runs on save when enabled |
| Import HTML/CSV (migration) | URLs land first; archive on demand or re-save important links |
Pinboard API posts/add | Same — archive policy per your account settings |
After Safari or X archive import, expect link-first rows until you archive — honest migration UX, not instant full-text magic.
Using search in the app
- Library search — keyword over indexed fields (fast filter bar).
- Ask / semantic flows — when enabled on your plan, query by intent over embedded chunks.
- Reader highlights — searchable via export and in-reader panel; library-wide Markdown export for PKM.
We deploy with FTS health checks in the production gate so empty vectors do not ship silently (see schema drift deploy gate).
vs “just use Google site:search”
Site search only works if the page is still live and unchanged. Cowpin search runs on what you captured — the reason people left Pocket/Omnivore was not only shutdown, but link rot.
vs AI auto-tag tools
Auto-tags help browsing; they do not replace literal phrase search in archived body text. Cowpin keeps tags you choose plus full-text — manual tags vs magic tags.
Agents
MCP/API search tools read the same library — search quality is a product promise, not a demo-only feature.
Practical habit
After any big import: pick ten critical links → open → confirm archive → search a unique phrase. That is your second-save for search, not just for habit (playbook).