System

Flatpak App Cache

Flatpak stores per-application data under ~/.var/app, where many sandboxed apps accumulate disposable cache files such as browser HTTP caches, thumbnail stores, Mesa and WebKit GPU shader caches, temporary download fragments, and SQLite WAL or journal files created during normal use. These caches exist to speed up launches, page loads, media decoding, and database writes inside the sandbox, but they can grow large or become mismatched after app, runtime, or graphics driver updates. Kudu removes only the cache and temporary data from Flatpak app directories, leaving documents, saved sessions, accounts, settings, and other user data untouched.

Why clean Flatpak App Cache?

  • Stale Mesa or WebKit shader blobs after a graphics driver or Flatpak runtime update can cause black windows, flickering, or heavy stutter until the GPU cache is rebuilt
  • Oversized browser-style HTTP caches inside sandboxed apps waste gigabytes in ~/.var/app and show up as unexpectedly high home-folder disk usage
  • Corrupted thumbnail and image caches lead to missing previews, blank icons, or avatars that reappear only after the app regenerates them
  • Leftover temporary downloads and crash-recovery files can make a Flatpak app reopen old tabs, prompts, or partial transfers that should have disappeared
  • SQLite databases used for search indexes, history, or local metadata can become page-fragmented over time, and VACUUM rewrites them compactly without deleting rows
  • Cache files built against an older Flatpak runtime may trigger slow first launches, repeated asset revalidation, or odd UI glitches after the app updates
  • Accumulated WAL and journal files from sandboxed databases can keep cache directories larger than expected and may coincide with slower startup while the app replays pending changes
What gets cleaned

Cache paths Kudu targets

Linux

~/.var/app
Frequently asked

Common questions about Flatpak App Cache

Free & open source

Download Kudu and reclaim your disk space.

Available on Windows, macOS, and Linux. No account required, no feature gates, no telemetry without consent. All cleaning targets are open source and community-auditable.