Find a file by name

Agents: claude, codex, opencode · On: phone, desktop

Type part of a name into the explorer's filter and the whole folder is walked for matches. The tree steps aside for a flat list of paths.

Why you would use it

You know the file is called something like runner but not which folder it sits in, and opening folders one at a time on a phone is slow.

Open the explorer (#files-open).
Type into the filter box (#explorer-filter, placeholder filter…).
Tap a file result to open it in the viewer.
Tap a folder result to show it in the tree: the filter clears, every folder on the way down opens, and the row is scrolled into the middle of the panel.
Press Escape, or clear the box, to go back to the tree. With the box already empty, Escape in the drawer closes the drawer; in the docked column it does nothing.

Each result is one row showing its whole path from the thread's folder, not just its name. A file result carries its icon, and a changed or ignored one its letter and colour, as in the tree.
No match shows no match.
More than 500 matches ends in a Load more row that appends the next page.
The tree is not drawn while the filter has text in it.

Option	Default	What it changes
Typing delay	150 ms	How long typing stops before the server is asked. Every keystroke would otherwise walk the tree
Page	500 matches	How many come back at once
Depth	12 levels	How far below the folder the walk goes

It matches names, not file contents.
node_modules and .git are never walked into, though both are listed as ordinary folders in the tree.
Symlinked folders are not walked through: a cycle would never end, and what they point at is reachable by its own path.
The match is a plain substring, case insensitive. No globs, no regular expressions.
A folder that cannot be read is skipped without failing the search.
total for a query is the capped count of what the walk found, not an exact count of every match in the tree.