Should you replace your current file manager?
Try Flea alongside your current file manager if you already use Omarchy and want its keyboard controls and graphical previews. Keep your existing default until you have checked file operations, previews and any devices or shares you depend on. Our sandbox result supports a source-build assessment, not a completed desktop replacement.
Flea requires Omarchy and Quickshell; it will not install on plain Arch. The installation guide makes that restriction explicit. Choosing Flea therefore starts with your desktop, rather than a general comparison of Linux file managers.
What our September run established
Our run on September 28, 2026 used commit f304c38, identified by the release page as the v0.3.5 source revision. In our sandbox, the install step succeeded in 7 seconds and the build step succeeded in 15 seconds. The test command exited 101 after 39 seconds.
The harness records step success, exit codes, elapsed wall-clock time and parsed test summaries. Its stored parser summaries report 1,744 passed, 4 failed and 1,748 total. Those values remain the historical record. The attached raw output repeats a Cargo summary around a replay marker, so the aggregates may double-count suites and cannot establish unique or exhaustive test counts. A pass percentage would conceal that uncertainty.
The raw tail names a_file_carries_its_owner_and_the_line_puts_it_on_the_wire_escaped and a_path_resolves_through_its_own_uid_and_a_missing_path_resolves_to_nothing. Both panic with the message "the test runner's uid is a local account on this box". These are the individually visible failures, not a replacement total for the stored parser summaries. The message points to an account-resolution assumption in the sandbox, but it does not establish whether the cause belongs to the environment or the application.
The record contains no desktop interaction results. It does not verify opening the Quickshell interface, browsing SMB, connecting a phone, completing a file-chooser request or installing the Arch package. A zero exit code for the build does not demonstrate those workflows. No audit result was recorded, and scan coverage is unknown.
The MrKeyoor testing methodology explains how command outcomes and parser summaries reach a review. That page describes the publisher's current procedure; it does not prove that its current version governed this September run. The dated record above is the evidence for this checkout.
Keep v0.3.5 separate from current documentation
The README pinned to our tested commit describes list, column and grid views, tabs, Quick Look, file operations with an undo journal, network locations and connected devices. It also says rows and thumbnails load as needed. These are documented capabilities, not workflows measured by our harness.
The current README describes additional features, including a two-pane view. Those features must not be credited to our v0.3.5 run. Likewise, the project's published GUI benchmarks concern separate workloads and source revisions. Our install and build timings do not measure browsing speed or support a fastest-file-manager claim.
The v0.3.5 release notes identify a specific limitation: files dropped on Chromium's New Tab Page are refused, while google.com accepts them. That is a release-specific documented issue, not a result from our sandbox or proof of current behavior.
Try it before changing desktop defaults
The current installation guide says Flea is available only from the Omarchy Package Repository. Its earlier AUR packages are historical installation routes, not the current recommendation. Follow the current guide for a new installation rather than treating v0.3.5's package advice as current.
Installing Flea alone leaves default selection as a separate choice. The documented default command changes per-user MIME handling, the D-Bus registration used for "Show in folder", Hyprland bindings and portal preferences. File-chooser routing can also be selected separately.
This matters when removing the package. Pacman does not remove those per-user preferences. The guide documents flea --default off and flea --picker off; run the relevant undo commands before removing Flea. Otherwise, a leftover D-Bus service can still point to a removed executable.
For an evaluation, install without changing defaults, open a folder containing disposable copies, and check the operations you need. Test your actual shares, devices and file dialogs before routing everyday work through Flea. These are recommended checks, not completed parts of our run.
Alternatives depend on your desktop
Nautilus is GNOME's Files app. Choose it when you want the file browser developed for GNOME. Flea's installation guide also identifies Nautilus as the stock Omarchy directory handler, so retaining it while evaluating Flea is a practical fallback. This is a replacement decision for folder browsing; neither source establishes that Flea reproduces every Nautilus workflow.
Dolphin is KDE's file manager. Choose it when you want KDE's file manager for browsing drives and managing files. Its own documentation emphasizes discoverable features and usability. The supplied evidence does not establish a speed or compatibility advantage over Flea, so desktop fit is the useful distinction here.
How the editorial scores work
These are editorial assessments, not benchmark results or adoption measurements. The rubric runs from 1 to 5: 1 means substantial obstacles or little supporting evidence, 3 means a usable evaluation path with material constraints, and 5 means unusually complete supporting evidence for that dimension. Scores of 2 and 4 fall between those anchors.
Setup ease is 3/5: our install and build steps succeeded, but an actual desktop setup requires Omarchy, Quickshell and integration work outside the sandbox result. Documentation is 5/5 for its explicit dependencies, package instructions, integration behavior and removal guidance; that score does not certify every instruction through execution.
Community is 3/5, narrowly based on the release's credited outside contributors and the repositories' contribution routes. It does not measure adoption, response times or maintainer capacity. Maturity is 2/5 for this v0.3.5 assessment: the recorded test command failed and the desktop workflows were not verified. That rating does not assess later releases.