Microsoft Edge WebView2 Caches
Many Windows apps embed Microsoft Edge WebView2 to render web-based interfaces, and the runtime stores temporary data under Local AppData such as HTTP disk cache entries, compiled V8 code cache, GPUCache files, WebGL and shader blobs, and custom dictionary cache files. These caches exist to speed up page loads, JavaScript startup, text input, and GPU rendering inside embedded app windows, but they can become stale after app, runtime, or graphics driver changes. Kudu removes only those WebView2 cache directories and leaves cookies, signed-in sessions, IndexedDB, Local Storage, service-worker storage, offline content, settings, and saved account data alone.
Why clean Microsoft Edge WebView2 Caches?
- Corrupted HTTP disk cache entries can make an embedded login page or settings screen load with missing CSS, broken icons, or endless spinners until the cache is rebuilt
- Invalidated V8 code cache after a WebView2 runtime or app update often causes a slow first launch, frozen UI panels, or script errors in embedded web views
- Stale GPUCache data after a graphics driver update can break hardware-accelerated rendering, showing black panels, flicker, or blank embedded windows until fresh cache files are generated
- Outdated WebGL shader blobs from a previous runtime or driver version can cause stuttering, missing effects, or a white screen in apps that use canvas or 3D content inside WebView2
- An oversized cache tree under Local AppData can consume hundreds of megabytes across multiple apps, and users usually notice it as unexplained disk usage even though documents and app data are unchanged
- Damaged dictionary cache files can interfere with spellcheck suggestions in embedded editors, leading to missing corrections or delayed typing feedback while the dictionary cache repopulates
Cache paths Kudu targets
Windows
%LocalAppData% |
Common questions about Microsoft Edge WebView2 Caches
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.