Pagefind
Site search with no search server to run.
Try it. Nothing is listening.
This box searches all 20 posts on this site. The index was built when the page was, and every keystroke is answered by your own browser — open the network tab and watch nothing happen.
Nothing matches that. Which is also an answer the server never had to give.
How it works
Five steps, in the order they happen.
- 01
Turn on a view mode
Enable the Pagefind Index view mode on the content types you want found. It is the same Drupal furniture you already know.
- 02
Tag the fields
The Pagefind Expert formatter configures search behaviour directly from Manage display, so what gets indexed is a display decision, not a config file.
- 03
Build the index
From the dashboard, or with Drush once you are past a few thousand documents. What comes out is a set of static, compressed files.
- 04
The browser does the searching
Pagefind queries that index on the visitor’s own device via WebAssembly. There is no search server to provision, patch, scale, or pay for.
- 05
Then shape the results
Faceted listings in list, grid or table layouts. Featured results to promote a page for a keyword, and index-time synonym expansion for the words your visitors actually use.
The facts
Everything below is verifiable. Nothing is rounded up.
- Engine
- Pagefind, Rust-based, WebAssembly
- Search server
- None required
- Drupal
- 10.3+, 11
- PHP
- 8.3+
- Requires
- Private file system
- Designed for
- ~50,000 indexed documents
- Use Drush above
- 5,000 documents
- Extras
- Multisite aggregation, per-node preview tab
- Latest release
- 1.0.0-alpha2 — 26 June 2026